Last Year

2019
[1]
M. Abouhamra, “AligNarr: Aligning Narratives of Different Length for Movie Summarization,” Universität des Saarlandes, Saarbrücken, 2019.
Abstract
Automatic text alignment is an important problem in natural language processing. It can be used to create the data needed to train different language models. Most research about automatic summarization revolves around summarizing news articles or scientific papers, which are somewhat small texts with simple and clear structure. The bigger the difference in size between the summary and the original text, the harder the problem will be since important information will be sparser and identifying them can be more difficult. Therefore, creating datasets from larger texts can help improve automatic summarization. In this project, we try to develop an algorithm which can automatically create a dataset for abstractive automatic summarization for bigger narrative text bodies such as movie scripts. To this end, we chose sentences as summary text units and scenes as script text units and developed an algorithm which uses some of the latest natural language processing techniques to align scenes and sentences based on the similarity in their meanings. Solving this alignment problem can provide us with important information about how to evaluate the meaning of a text, which can help us create better abstractive summariza- tion models. We developed a method which uses different similarity scoring techniques (embedding similarity, word inclusion and entity inclusion) to align script scenes and sum- mary sentences which achieved an F1 score of 0.39. Analyzing our results showed that the bigger the differences in the number of text units being aligned, the more difficult the alignment problem is. We also critiqued of our own similarity scoring techniques and dif- ferent alignment algorithms based on integer linear programming and local optimization and showed their limitations and discussed ideas to improve them.
Export
BibTeX
@mastersthesis{AbouhamraMSc2019, TITLE = {{AligNarr}: Aligning Narratives of Different Length for Movie Summarization}, AUTHOR = {Abouhamra, Mostafa}, LANGUAGE = {eng}, SCHOOL = {Universit{\"a}t des Saarlandes}, ADDRESS = {Saarbr{\"u}cken}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, ABSTRACT = {Automatic text alignment is an important problem in natural language processing. It can be used to create the data needed to train different language models. Most research about automatic summarization revolves around summarizing news articles or scientific papers, which are somewhat small texts with simple and clear structure. The bigger the difference in size between the summary and the original text, the harder the problem will be since important information will be sparser and identifying them can be more difficult. Therefore, creating datasets from larger texts can help improve automatic summarization. In this project, we try to develop an algorithm which can automatically create a dataset for abstractive automatic summarization for bigger narrative text bodies such as movie scripts. To this end, we chose sentences as summary text units and scenes as script text units and developed an algorithm which uses some of the latest natural language processing techniques to align scenes and sentences based on the similarity in their meanings. Solving this alignment problem can provide us with important information about how to evaluate the meaning of a text, which can help us create better abstractive summariza- tion models. We developed a method which uses different similarity scoring techniques (embedding similarity, word inclusion and entity inclusion) to align script scenes and sum- mary sentences which achieved an F1 score of 0.39. Analyzing our results showed that the bigger the differences in the number of text units being aligned, the more difficult the alignment problem is. We also critiqued of our own similarity scoring techniques and dif- ferent alignment algorithms based on integer linear programming and local optimization and showed their limitations and discussed ideas to improve them.}, }
Endnote
%0 Thesis %A Abouhamra, Mostafa %Y Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society International Max Planck Research School, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T AligNarr: Aligning Narratives of Different Length for Movie Summarization : %G eng %U http://hdl.handle.net/21.11116/0000-0004-5836-D %I Universität des Saarlandes %C Saarbrücken %D 2019 %P 54 p. %V master %9 master %X Automatic text alignment is an important problem in natural language processing. It can be used to create the data needed to train different language models. Most research about automatic summarization revolves around summarizing news articles or scientific papers, which are somewhat small texts with simple and clear structure. The bigger the difference in size between the summary and the original text, the harder the problem will be since important information will be sparser and identifying them can be more difficult. Therefore, creating datasets from larger texts can help improve automatic summarization. In this project, we try to develop an algorithm which can automatically create a dataset for abstractive automatic summarization for bigger narrative text bodies such as movie scripts. To this end, we chose sentences as summary text units and scenes as script text units and developed an algorithm which uses some of the latest natural language processing techniques to align scenes and sentences based on the similarity in their meanings. Solving this alignment problem can provide us with important information about how to evaluate the meaning of a text, which can help us create better abstractive summariza- tion models. We developed a method which uses different similarity scoring techniques (embedding similarity, word inclusion and entity inclusion) to align script scenes and sum- mary sentences which achieved an F1 score of 0.39. Analyzing our results showed that the bigger the differences in the number of text units being aligned, the more difficult the alignment problem is. We also critiqued of our own similarity scoring techniques and dif- ferent alignment algorithms based on integer linear programming and local optimization and showed their limitations and discussed ideas to improve them.
[2]
A. Abujabal, “Question Answering over Knowledge Bases with Continuous Learning,” Universität des Saarlandes, Saarbrücken, 2019.
Abstract
Answering complex natural language questions with crisp answers is crucial towards satisfying the information needs of advanced users. With the rapid growth of knowledge bases (KBs) such as Yago and Freebase, this goal has become attainable by translating questions into formal queries like SPARQL queries. Such queries can then be evaluated over knowledge bases to retrieve crisp answers. To this end, three research issues arise: (i) how to develop methods that are robust to lexical and syntactic variations in questions and can handle complex questions, (ii) how to design and curate datasets to advance research in question answering, and (iii) how to efficiently identify named entities in questions. In this dissertation, we make the following five contributions in the areas of question answering (QA) and named entity recognition (NER). For issue (i), we make the following contributions: We present QUINT, an approach for answering natural language questions over knowledge bases using automatically learned templates. Templates are an important asset for QA over KBs, simplifying the semantic parsing of input questions and generating formal queries for interpretable answers. QUINT is capable of answering both simple and compositional questions. We introduce NEQA, a framework for continuous learning for QA over KBs. NEQA starts with a small seed of training examples in the form of question-answer pairs, and improves its performance over time. NEQA combines both syntax, through template-based answering, and semantics, via a semantic similarity function. %when templates fail to do so. Moreover, it adapts to the language used after deployment by periodically retraining its underlying models. For issues (i) and (ii), we present TEQUILA, a framework for answering complex questions with explicit and implicit temporal conditions over KBs. TEQUILA is built on a rule-based framework that detects and decomposes temporal questions into simpler sub-questions that can be answered by standard KB-QA systems. TEQUILA reconciles the results of sub-questions into final answers. TEQUILA is accompanied with a dataset called TempQuestions, which consists of 1,271 temporal questions with gold-standard answers over Freebase. This collection is derived by judiciously selecting time-related questions from existing QA datasets. For issue (ii), we publish ComQA, a large-scale manually-curated dataset for QA. ComQA contains questions that represent real information needs and exhibit a wide range of difficulties such as the need for temporal reasoning, comparison, and compositionality. ComQA contains paraphrase clusters of semantically-equivalent questions that can be exploited by QA systems. We harness a combination of community question-answering platforms and crowdsourcing to construct the ComQA dataset. For issue (iii), we introduce a neural network model based on subword units for named entity recognition. The model learns word representations using a combination of characters, bytes and phonemes. While achieving comparable performance with word-level based models, our model has an order-of-magnitude smaller vocabulary size and lower memory requirements, and it handles out-of-vocabulary words.
Export
BibTeX
@phdthesis{Abujabalphd2013, TITLE = {Question Answering over Knowledge Bases with Continuous Learning}, AUTHOR = {Abujabal, Abdalghani}, LANGUAGE = {eng}, DOI = {10.22028/D291-27968}, SCHOOL = {Universit{\"a}t des Saarlandes}, ADDRESS = {Saarbr{\"u}cken}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, ABSTRACT = {Answering complex natural language questions with crisp answers is crucial towards satisfying the information needs of advanced users. With the rapid growth of knowledge bases (KBs) such as Yago and Freebase, this goal has become attainable by translating questions into formal queries like SPARQL queries. Such queries can then be evaluated over knowledge bases to retrieve crisp answers. To this end, three research issues arise: (i) how to develop methods that are robust to lexical and syntactic variations in questions and can handle complex questions, (ii) how to design and curate datasets to advance research in question answering, and (iii) how to efficiently identify named entities in questions. In this dissertation, we make the following five contributions in the areas of question answering (QA) and named entity recognition (NER). For issue (i), we make the following contributions: We present QUINT, an approach for answering natural language questions over knowledge bases using automatically learned templates. Templates are an important asset for QA over KBs, simplifying the semantic parsing of input questions and generating formal queries for interpretable answers. QUINT is capable of answering both simple and compositional questions. We introduce NEQA, a framework for continuous learning for QA over KBs. NEQA starts with a small seed of training examples in the form of question-answer pairs, and improves its performance over time. NEQA combines both syntax, through template-based answering, and semantics, via a semantic similarity function. %when templates fail to do so. Moreover, it adapts to the language used after deployment by periodically retraining its underlying models. For issues (i) and (ii), we present TEQUILA, a framework for answering complex questions with explicit and implicit temporal conditions over KBs. TEQUILA is built on a rule-based framework that detects and decomposes temporal questions into simpler sub-questions that can be answered by standard KB-QA systems. TEQUILA reconciles the results of sub-questions into final answers. TEQUILA is accompanied with a dataset called TempQuestions, which consists of 1,271 temporal questions with gold-standard answers over Freebase. This collection is derived by judiciously selecting time-related questions from existing QA datasets. For issue (ii), we publish ComQA, a large-scale manually-curated dataset for QA. ComQA contains questions that represent real information needs and exhibit a wide range of difficulties such as the need for temporal reasoning, comparison, and compositionality. ComQA contains paraphrase clusters of semantically-equivalent questions that can be exploited by QA systems. We harness a combination of community question-answering platforms and crowdsourcing to construct the ComQA dataset. For issue (iii), we introduce a neural network model based on subword units for named entity recognition. The model learns word representations using a combination of characters, bytes and phonemes. While achieving comparable performance with word-level based models, our model has an order-of-magnitude smaller vocabulary size and lower memory requirements, and it handles out-of-vocabulary words.}, }
Endnote
%0 Thesis %A Abujabal, Abdalghani %Y Weikum, Gerhard %A referee: Linn, Jimmy %A referee: Berberich, Klaus %+ Databases and Information Systems, MPI for Informatics, Max Planck Society International Max Planck Research School, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society %T Question Answering over Knowledge Bases with Continuous Learning : %G eng %U http://hdl.handle.net/21.11116/0000-0003-AEC0-0 %R 10.22028/D291-27968 %I Universität des Saarlandes %C Saarbrücken %D 2019 %P 141 p. %V phd %9 phd %X Answering complex natural language questions with crisp answers is crucial towards satisfying the information needs of advanced users. With the rapid growth of knowledge bases (KBs) such as Yago and Freebase, this goal has become attainable by translating questions into formal queries like SPARQL queries. Such queries can then be evaluated over knowledge bases to retrieve crisp answers. To this end, three research issues arise: (i) how to develop methods that are robust to lexical and syntactic variations in questions and can handle complex questions, (ii) how to design and curate datasets to advance research in question answering, and (iii) how to efficiently identify named entities in questions. In this dissertation, we make the following five contributions in the areas of question answering (QA) and named entity recognition (NER). For issue (i), we make the following contributions: We present QUINT, an approach for answering natural language questions over knowledge bases using automatically learned templates. Templates are an important asset for QA over KBs, simplifying the semantic parsing of input questions and generating formal queries for interpretable answers. QUINT is capable of answering both simple and compositional questions. We introduce NEQA, a framework for continuous learning for QA over KBs. NEQA starts with a small seed of training examples in the form of question-answer pairs, and improves its performance over time. NEQA combines both syntax, through template-based answering, and semantics, via a semantic similarity function. %when templates fail to do so. Moreover, it adapts to the language used after deployment by periodically retraining its underlying models. For issues (i) and (ii), we present TEQUILA, a framework for answering complex questions with explicit and implicit temporal conditions over KBs. TEQUILA is built on a rule-based framework that detects and decomposes temporal questions into simpler sub-questions that can be answered by standard KB-QA systems. TEQUILA reconciles the results of sub-questions into final answers. TEQUILA is accompanied with a dataset called TempQuestions, which consists of 1,271 temporal questions with gold-standard answers over Freebase. This collection is derived by judiciously selecting time-related questions from existing QA datasets. For issue (ii), we publish ComQA, a large-scale manually-curated dataset for QA. ComQA contains questions that represent real information needs and exhibit a wide range of difficulties such as the need for temporal reasoning, comparison, and compositionality. ComQA contains paraphrase clusters of semantically-equivalent questions that can be exploited by QA systems. We harness a combination of community question-answering platforms and crowdsourcing to construct the ComQA dataset. For issue (iii), we introduce a neural network model based on subword units for named entity recognition. The model learns word representations using a combination of characters, bytes and phonemes. While achieving comparable performance with word-level based models, our model has an order-of-magnitude smaller vocabulary size and lower memory requirements, and it handles out-of-vocabulary words. %U https://publikationen.sulb.uni-saarland.de/handle/20.500.11880/27438
[3]
A. Abujabal, R. Saha Roy, M. Yahya, and G. Weikum, “ComQA: A Community-sourced Dataset for Complex Factoid Question Answering with Paraphrase Clusters,” in The 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL HLT 2019), Minneapolis, MN, USA, 2019.
Export
BibTeX
@inproceedings{abujabal19comqa, TITLE = {{ComQA}: {A} Community-sourced Dataset for Complex Factoid Question Answering with Paraphrase Clusters}, AUTHOR = {Abujabal, Abdalghani and Saha Roy, Rishiraj and Yahya, Mohamed and Weikum, Gerhard}, LANGUAGE = {eng}, ISBN = {978-1-950737-13-0}, URL = {https://www.aclweb.org/anthology/N19-1027}, PUBLISHER = {ACL}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {The 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL HLT 2019)}, EDITOR = {Burstein, Jill and Doran, Christy and Solorio, Thamar}, PAGES = {307--317}, ADDRESS = {Minneapolis, MN, USA}, }
Endnote
%0 Conference Proceedings %A Abujabal, Abdalghani %A Saha Roy, Rishiraj %A Yahya, Mohamed %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society %T ComQA: A Community-sourced Dataset for Complex Factoid Question Answering with Paraphrase Clusters : %G eng %U http://hdl.handle.net/21.11116/0000-0003-11A7-D %U https://www.aclweb.org/anthology/N19-1027 %D 2019 %B Annual Conference of the North American Chapter of the Association for Computational Linguistics %Z date of event: 2019-06-02 - 2019-06-07 %C Minneapolis, MN, USA %B The 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies %E Burstein, Jill; Doran, Christy; Solorio, Thamar %P 307 - 317 %I ACL %@ 978-1-950737-13-0 %U https://www.aclweb.org/anthology/N19-1027
[4]
M. Alikhani, S. Nag Chowdhury, G. de Melo, and M. Stone, “CITE: A Corpus Of Text-Image Discourse Relations,” in Proceedings of the 2019 Annual Conference of the North American Chapter of the Association for Computational Linguistics (NAACL-HLT 2019), Minneapolis, MN, USA, 2019.
Export
BibTeX
@inproceedings{AlikhaniEtAl2019CITETextImageDiscourse, TITLE = {{CITE}: {A} Corpus Of Text-Image Discourse Relations}, AUTHOR = {Alikhani, Malihe and Nag Chowdhury, Sreyasi and de Melo, Gerard and Stone, Matthew}, LANGUAGE = {eng}, ISBN = {978-1-950737-13-0}, PUBLISHER = {ACL}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {Proceedings of the 2019 Annual Conference of the North American Chapter of the Association for Computational Linguistics (NAACL-HLT 2019)}, EDITOR = {Burstein, Jill and Doran, Christy and Solorio, Thamar}, PAGES = {570--575}, ADDRESS = {Minneapolis, MN, USA}, }
Endnote
%0 Conference Proceedings %A Alikhani, Malihe %A Nag Chowdhury, Sreyasi %A de Melo, Gerard %A Stone, Matthew %+ External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations %T CITE: A Corpus Of Text-Image Discourse Relations : %G eng %U http://hdl.handle.net/21.11116/0000-0003-78D8-3 %D 2019 %B Annual Conference of the North American Chapter of the Association for Computational Linguistics %Z date of event: 2019-06-02 - 2019-06-07 %C Minneapolis, MN, USA %B Proceedings of the 2019 Annual Conference of the North American Chapter of the Association for Computational Linguistics %E Burstein, Jill; Doran, Christy; Solorio, Thamar %P 570 - 575 %I ACL %@ 978-1-950737-13-0 %U https://aclweb.org/anthology/papers/N/N19/N19-1056/
[5]
S. Arora and A. Yates, “Investigating Retrieval Method Selection with Axiomatic Features,” in Proceedings of the 1st Interdisciplinary Workshop on Algorithm Selection and Meta-Learning in Information Retrieval co-located with the 41st European Conference on Information Retrieval (ECIR 2019) (AMIR 2019), Cologne, Germany, 2019.
Export
BibTeX
@inproceedings{Arora_AMIR2019, TITLE = {Investigating Retrieval Method Selection with Axiomatic Features}, AUTHOR = {Arora, Siddhant and Yates, Andrew}, LANGUAGE = {eng}, ISSN = {1613-0073}, URL = {urn:nbn:de:0074-2360-3}, PUBLISHER = {CEUR-WS}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {Proceedings of the 1st Interdisciplinary Workshop on Algorithm Selection and Meta-Learning in Information Retrieval co-located with the 41st European Conference on Information Retrieval (ECIR 2019) (AMIR 2019)}, EDITOR = {Beel, Joeran and Kolthoff, Lars}, PAGES = {18--31}, EID = {4}, SERIES = {CEUR Workshop Proceedings}, VOLUME = {2360}, ADDRESS = {Cologne, Germany}, }
Endnote
%0 Conference Proceedings %A Arora, Siddhant %A Yates, Andrew %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T Investigating Retrieval Method Selection with Axiomatic Features : %G eng %U http://hdl.handle.net/21.11116/0000-0004-028E-A %D 2019 %B The 1st Interdisciplinary Workshop on Algorithm Selection and Meta-Learning in Information Retrieval %Z date of event: 2019-04-14 - 2019-04-14 %C Cologne, Germany %B Proceedings of the 1st Interdisciplinary Workshop on Algorithm Selection and Meta-Learning in Information Retrieval co-located with the 41st European Conference on Information Retrieval (ECIR 2019) %E Beel, Joeran; Kolthoff, Lars %P 18 - 31 %Z sequence number: 4 %I CEUR-WS %B CEUR Workshop Proceedings %N 2360 %@ false %U http://ceur-ws.org/Vol-2360/paper4Axiomatic.pdf
[6]
S. Arora and A. Yates, “Investigating Retrieval Method Selection with Axiomatic Features,” 2019. [Online]. Available: http://arxiv.org/abs/1904.05737. (arXiv: 1904.05737)
Abstract
We consider algorithm selection in the context of ad-hoc information retrieval. Given a query and a pair of retrieval methods, we propose a meta-learner that predicts how to combine the methods' relevance scores into an overall relevance score. Inspired by neural models' different properties with regard to IR axioms, these predictions are based on features that quantify axiom-related properties of the query and its top ranked documents. We conduct an evaluation on TREC Web Track data and find that the meta-learner often significantly improves over the individual methods. Finally, we conduct feature and query weight analyses to investigate the meta-learner's behavior.
Export
BibTeX
@online{Arora_arXiv1904.05737, TITLE = {Investigating Retrieval Method Selection with Axiomatic Features}, AUTHOR = {Arora, Siddhant and Yates, Andrew}, LANGUAGE = {eng}, URL = {http://arxiv.org/abs/1904.05737}, EPRINT = {1904.05737}, EPRINTTYPE = {arXiv}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, ABSTRACT = {We consider algorithm selection in the context of ad-hoc information retrieval. Given a query and a pair of retrieval methods, we propose a meta-learner that predicts how to combine the methods' relevance scores into an overall relevance score. Inspired by neural models' different properties with regard to IR axioms, these predictions are based on features that quantify axiom-related properties of the query and its top ranked documents. We conduct an evaluation on TREC Web Track data and find that the meta-learner often significantly improves over the individual methods. Finally, we conduct feature and query weight analyses to investigate the meta-learner's behavior.}, }
Endnote
%0 Report %A Arora, Siddhant %A Yates, Andrew %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T Investigating Retrieval Method Selection with Axiomatic Features : %G eng %U http://hdl.handle.net/21.11116/0000-0004-02BF-3 %U http://arxiv.org/abs/1904.05737 %D 2019 %X We consider algorithm selection in the context of ad-hoc information retrieval. Given a query and a pair of retrieval methods, we propose a meta-learner that predicts how to combine the methods' relevance scores into an overall relevance score. Inspired by neural models' different properties with regard to IR axioms, these predictions are based on features that quantify axiom-related properties of the query and its top ranked documents. We conduct an evaluation on TREC Web Track data and find that the meta-learner often significantly improves over the individual methods. Finally, we conduct feature and query weight analyses to investigate the meta-learner's behavior. %K Computer Science, Information Retrieval, cs.IR
[7]
A. Chakraborty, N. Mota, A. J. Biega, K. P. Gummadi, and H. Heidari, “On the Impact of Choice Architectures on Inequality in Online Donation Platforms,” in Proceedings of The World Wide Web Conference (WWW 2019), San Francisco, CA, USA, 2019.
Export
BibTeX
@inproceedings{Chakraborty_WWW2019, TITLE = {On the Impact of Choice Architectures on Inequality in Online Donation Platforms}, AUTHOR = {Chakraborty, Abhijnan and Mota, Nuno and Biega, Asia J. and Gummadi, Krishna P. and Heidari, Hoda}, LANGUAGE = {eng}, ISBN = {978-1-4503-6674-8}, DOI = {10.1145/3308558.3313663}, PUBLISHER = {ACM}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, BOOKTITLE = {Proceedings of The World Wide Web Conference (WWW 2019)}, PAGES = {2623--2629}, ADDRESS = {San Francisco, CA, USA}, }
Endnote
%0 Conference Proceedings %A Chakraborty, Abhijnan %A Mota, Nuno %A Biega, Asia J. %A Gummadi, Krishna P. %A Heidari, Hoda %+ External Organizations External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations %T On the Impact of Choice Architectures on Inequality in Online Donation Platforms : %G eng %U http://hdl.handle.net/21.11116/0000-0002-FC88-9 %R 10.1145/3308558.3313663 %D 2019 %B The Web Conference %Z date of event: 2019-05-13 - 2019-05-17 %C San Francisco, CA, USA %B Proceedings of The World Wide Web Conference %P 2623 - 2629 %I ACM %@ 978-1-4503-6674-8
[8]
F. Chierichetti, R. Kumar, A. Panconesi, and E. Terolli, “On the Distortion of Locality Sensitive Hashing,” SIAM Journal on Computing, vol. 48, no. 2, 2019.
Export
BibTeX
@article{Chierichetti2019, TITLE = {On the Distortion of Locality Sensitive Hashing}, AUTHOR = {Chierichetti, Flavio and Kumar, Ravi and Panconesi, Alessandro and Terolli, Erisa}, LANGUAGE = {eng}, ISSN = {0097-5397}, DOI = {10.1137/17M1127752}, PUBLISHER = {SIAM}, ADDRESS = {Philadelphia, PA}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, JOURNAL = {SIAM Journal on Computing}, VOLUME = {48}, NUMBER = {2}, PAGES = {350--372}, }
Endnote
%0 Journal Article %A Chierichetti, Flavio %A Kumar, Ravi %A Panconesi, Alessandro %A Terolli, Erisa %+ External Organizations External Organizations External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society %T On the Distortion of Locality Sensitive Hashing : %G eng %U http://hdl.handle.net/21.11116/0000-0003-A7E7-C %R 10.1137/17M1127752 %7 2019 %D 2019 %J SIAM Journal on Computing %V 48 %N 2 %& 350 %P 350 - 372 %I SIAM %C Philadelphia, PA %@ false
[9]
P. Christmann, R. Saha Roy, A. Abujabal, J. Singh, and G. Weikum, “Look before you Hop: Conversational Question Answering over Knowledge Graphs Using Judicious Context Expansion,” 2019. [Online]. Available: http://arxiv.org/abs/1910.03262. (arXiv: 1910.03262)
Abstract
Fact-centric information needs are rarely one-shot; users typically ask follow-up questions to explore a topic. In such a conversational setting, the user's inputs are often incomplete, with entities or predicates left out, and ungrammatical phrases. This poses a huge challenge to question answering (QA) systems that typically rely on cues in full-fledged interrogative sentences. As a solution, we develop CONVEX: an unsupervised method that can answer incomplete questions over a knowledge graph (KG) by maintaining conversation context using entities and predicates seen so far and automatically inferring missing or ambiguous pieces for follow-up questions. The core of our method is a graph exploration algorithm that judiciously expands a frontier to find candidate answers for the current question. To evaluate CONVEX, we release ConvQuestions, a crowdsourced benchmark with 11,200 distinct conversations from five different domains. We show that CONVEX: (i) adds conversational support to any stand-alone QA system, and (ii) outperforms state-of-the-art baselines and question completion strategies.
Export
BibTeX
@online{Christmann_arXiv1910.03262, TITLE = {Look before you Hop: Conversational Question Answering over Knowledge Graphs Using Judicious Context Expansion}, AUTHOR = {Christmann, Phlipp and Saha Roy, Rishiraj and Abujabal, Abdalghani and Singh, Jyotsna and Weikum, Gerhard}, LANGUAGE = {eng}, URL = {http://arxiv.org/abs/1910.03262}, EPRINT = {1910.03262}, EPRINTTYPE = {arXiv}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, ABSTRACT = {Fact-centric information needs are rarely one-shot; users typically ask follow-up questions to explore a topic. In such a conversational setting, the user's inputs are often incomplete, with entities or predicates left out, and ungrammatical phrases. This poses a huge challenge to question answering (QA) systems that typically rely on cues in full-fledged interrogative sentences. As a solution, we develop CONVEX: an unsupervised method that can answer incomplete questions over a knowledge graph (KG) by maintaining conversation context using entities and predicates seen so far and automatically inferring missing or ambiguous pieces for follow-up questions. The core of our method is a graph exploration algorithm that judiciously expands a frontier to find candidate answers for the current question. To evaluate CONVEX, we release ConvQuestions, a crowdsourced benchmark with 11,200 distinct conversations from five different domains. We show that CONVEX: (i) adds conversational support to any stand-alone QA system, and (ii) outperforms state-of-the-art baselines and question completion strategies.}, }
Endnote
%0 Report %A Christmann, Phlipp %A Saha Roy, Rishiraj %A Abujabal, Abdalghani %A Singh, Jyotsna %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T Look before you Hop: Conversational Question Answering over Knowledge Graphs Using Judicious Context Expansion : %G eng %U http://hdl.handle.net/21.11116/0000-0005-83DC-F %U http://arxiv.org/abs/1910.03262 %D 2019 %X Fact-centric information needs are rarely one-shot; users typically ask follow-up questions to explore a topic. In such a conversational setting, the user's inputs are often incomplete, with entities or predicates left out, and ungrammatical phrases. This poses a huge challenge to question answering (QA) systems that typically rely on cues in full-fledged interrogative sentences. As a solution, we develop CONVEX: an unsupervised method that can answer incomplete questions over a knowledge graph (KG) by maintaining conversation context using entities and predicates seen so far and automatically inferring missing or ambiguous pieces for follow-up questions. The core of our method is a graph exploration algorithm that judiciously expands a frontier to find candidate answers for the current question. To evaluate CONVEX, we release ConvQuestions, a crowdsourced benchmark with 11,200 distinct conversations from five different domains. We show that CONVEX: (i) adds conversational support to any stand-alone QA system, and (ii) outperforms state-of-the-art baselines and question completion strategies. %K Computer Science, Information Retrieval, cs.IR,Computer Science, Computation and Language, cs.CL
[10]
P. Christmann, R. Saha Roy, A. Abujabal, J. Singh, and G. Weikum, “Look before you Hop: Conversational Question Answering over Knowledge Graphs Using Judicious Context Expansion,” in CIKM’19, 28th ACM International Conference on Information and Knowledge Management, Beijing China, 2019.
Export
BibTeX
@inproceedings{Christmann_CIKM2019, TITLE = {Look before you Hop: {C}onversational Question Answering over Knowledge Graphs Using Judicious Context Expansion}, AUTHOR = {Christmann, Phlipp and Saha Roy, Rishiraj and Abujabal, Abdalghani and Singh, Jyotsna and Weikum, Gerhard}, LANGUAGE = {eng}, ISBN = {9781450369763}, DOI = {10.1145/3357384.3358016}, PUBLISHER = {ACM}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, BOOKTITLE = {CIKM'19, 28th ACM International Conference on Information and Knowledge Management}, EDITOR = {Zhu, Wenwu and Tao, Dacheng}, PAGES = {729--738}, ADDRESS = {Beijing China}, }
Endnote
%0 Conference Proceedings %A Christmann, Phlipp %A Saha Roy, Rishiraj %A Abujabal, Abdalghani %A Singh, Jyotsna %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T Look before you Hop: Conversational Question Answering over Knowledge Graphs Using Judicious Context Expansion : %G eng %U http://hdl.handle.net/21.11116/0000-0005-8231-0 %R 10.1145/3357384.3358016 %D 2019 %B 28th ACM International Conference on Information and Knowledge Management %Z date of event: 2019-11-03 - 2019-11-07 %C Beijing China %B CIKM'19 %E Zhu, Wenwu; Tao, Dacheng %P 729 - 738 %I ACM %@ 9781450369763
[11]
C. X. Chu, S. Razniewski, and G. Weikum, “TiFi: Taxonomy Induction for Fictional Domains [Extended version],” 2019. [Online]. Available: http://arxiv.org/abs/1901.10263. (arXiv: 1901.10263)
Abstract
Taxonomies are important building blocks of structured knowledge bases, and their construction from text sources and Wikipedia has received much attention. In this paper we focus on the construction of taxonomies for fictional domains, using noisy category systems from fan wikis or text extraction as input. Such fictional domains are archetypes of entity universes that are poorly covered by Wikipedia, such as also enterprise-specific knowledge bases or highly specialized verticals. Our fiction-targeted approach, called TiFi, consists of three phases: (i) category cleaning, by identifying candidate categories that truly represent classes in the domain of interest, (ii) edge cleaning, by selecting subcategory relationships that correspond to class subsumption, and (iii) top-level construction, by mapping classes onto a subset of high-level WordNet categories. A comprehensive evaluation shows that TiFi is able to construct taxonomies for a diverse range of fictional domains such as Lord of the Rings, The Simpsons or Greek Mythology with very high precision and that it outperforms state-of-the-art baselines for taxonomy induction by a substantial margin.
Export
BibTeX
@online{Chu_arXIv1901.10263, TITLE = {{TiFi}: Taxonomy Induction for Fictional Domains [Extended version]}, AUTHOR = {Chu, Cuong Xuan and Razniewski, Simon and Weikum, Gerhard}, LANGUAGE = {eng}, URL = {http://arxiv.org/abs/1901.10263}, EPRINT = {1901.10263}, EPRINTTYPE = {arXiv}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, ABSTRACT = {Taxonomies are important building blocks of structured knowledge bases, and their construction from text sources and Wikipedia has received much attention. In this paper we focus on the construction of taxonomies for fictional domains, using noisy category systems from fan wikis or text extraction as input. Such fictional domains are archetypes of entity universes that are poorly covered by Wikipedia, such as also enterprise-specific knowledge bases or highly specialized verticals. Our fiction-targeted approach, called TiFi, consists of three phases: (i) category cleaning, by identifying candidate categories that truly represent classes in the domain of interest, (ii) edge cleaning, by selecting subcategory relationships that correspond to class subsumption, and (iii) top-level construction, by mapping classes onto a subset of high-level WordNet categories. A comprehensive evaluation shows that TiFi is able to construct taxonomies for a diverse range of fictional domains such as Lord of the Rings, The Simpsons or Greek Mythology with very high precision and that it outperforms state-of-the-art baselines for taxonomy induction by a substantial margin.}, }
Endnote
%0 Report %A Chu, Cuong Xuan %A Razniewski, Simon %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T TiFi: Taxonomy Induction for Fictional Domains [Extended version] : %G eng %U http://hdl.handle.net/21.11116/0000-0003-FE67-C %U http://arxiv.org/abs/1901.10263 %D 2019 %X Taxonomies are important building blocks of structured knowledge bases, and their construction from text sources and Wikipedia has received much attention. In this paper we focus on the construction of taxonomies for fictional domains, using noisy category systems from fan wikis or text extraction as input. Such fictional domains are archetypes of entity universes that are poorly covered by Wikipedia, such as also enterprise-specific knowledge bases or highly specialized verticals. Our fiction-targeted approach, called TiFi, consists of three phases: (i) category cleaning, by identifying candidate categories that truly represent classes in the domain of interest, (ii) edge cleaning, by selecting subcategory relationships that correspond to class subsumption, and (iii) top-level construction, by mapping classes onto a subset of high-level WordNet categories. A comprehensive evaluation shows that TiFi is able to construct taxonomies for a diverse range of fictional domains such as Lord of the Rings, The Simpsons or Greek Mythology with very high precision and that it outperforms state-of-the-art baselines for taxonomy induction by a substantial margin. %K Computer Science, Computation and Language, cs.CL,Computer Science, Artificial Intelligence, cs.AI,Computer Science, Information Retrieval, cs.IR
[12]
C. X. Chu, S. Razniewski, and G. Weikum, “TiFi: Taxonomy Induction for Fictional Domains,” in Proceedings of The World Wide Web Conference (WWW 2019), San Francisco, CA, USA, 2019.
Export
BibTeX
@inproceedings{Chu_WWW2019, TITLE = {{TiFi}: {T}axonomy Induction for Fictional Domains}, AUTHOR = {Chu, Cuong Xuan and Razniewski, Simon and Weikum, Gerhard}, LANGUAGE = {eng}, ISBN = {978-1-4503-6674-8}, DOI = {10.1145/3308558.3313519}, PUBLISHER = {ACM}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {Proceedings of The World Wide Web Conference (WWW 2019)}, EDITOR = {McAuley, Julian}, PAGES = {2673--2679}, ADDRESS = {San Francisco, CA, USA}, }
Endnote
%0 Conference Proceedings %A Chu, Cuong Xuan %A Razniewski, Simon %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T TiFi: Taxonomy Induction for Fictional Domains : %G eng %U http://hdl.handle.net/21.11116/0000-0003-6558-9 %R 10.1145/3308558.3313519 %D 2019 %B The Web Conference %Z date of event: 2019-05-13 - 2019-05-17 %C San Francisco, CA, USA %B Proceedings of The World Wide Web Conference %E McAuley, Julian %P 2673 - 2679 %I ACM %@ 978-1-4503-6674-8
[13]
I. Dikeoulias, J. Strötgen, and S. Razniewski, “Epitaph or Breaking News? Analyzing and Predicting the Stability of Knowledge Base Properties,” in Companion of The World Wide Web Conference (WWW 2019), San Francisco, CA, USA, 2019.
Export
BibTeX
@inproceedings{Dikeoulias_WWW2019, TITLE = {Epitaph or Breaking News? {A}nalyzing and Predicting the Stability of Knowledge Base Properties}, AUTHOR = {Dikeoulias, Ioannis and Str{\"o}tgen, Jannik and Razniewski, Simon}, LANGUAGE = {eng}, ISBN = {978-1-4503-6675-5}, DOI = {10.1145/3308560.3314998}, PUBLISHER = {ACM}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {Companion of The World Wide Web Conference (WWW 2019)}, EDITOR = {McAuley, Julian}, PAGES = {1155--1158}, ADDRESS = {San Francisco, CA, USA}, }
Endnote
%0 Conference Proceedings %A Dikeoulias, Ioannis %A Strötgen, Jannik %A Razniewski, Simon %+ External Organizations External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society %T Epitaph or Breaking News? Analyzing and Predicting the Stability of Knowledge Base Properties : %G eng %U http://hdl.handle.net/21.11116/0000-0004-0281-7 %R 10.1145/3308560.3314998 %D 2019 %B The Web Conference %Z date of event: 2019-05-13 - 2019-05-17 %C San Francisco, CA, USA %B Companion of The World Wide Web Conference %E McAuley, Julian %P 1155 - 1158 %I ACM %@ 978-1-4503-6675-5
[14]
P. Ernst, E. Terolli, and G. Weikum, “LongLife: a Platform for Personalized Searchfor Health and Life Sciences,” in Proceedings of the ISWC 2019 Satellite Tracks (Posters & Demonstrations, Industry, and Outrageous Ideas) co-located with 18th International Semantic Web Conference (ISWC 2019) (ISWC 2019 Satellites), Auckland, New Zealand, 2019.
Export
BibTeX
@inproceedings{Ernst_ISWC2019, TITLE = {{LongLife}: a Platform for Personalized Searchfor Health and Life Sciences}, AUTHOR = {Ernst, Patrick and Terolli, Erisa and Weikum, Gerhard}, LANGUAGE = {eng}, ISSN = {1613-0073}, URL = {http://ceur-ws.org/Vol-2456/paper62.pdf; urn:nbn:de:0074-2456-4}, PUBLISHER = {ceur-ws.org}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {Proceedings of the ISWC 2019 Satellite Tracks (Posters \& Demonstrations, Industry, and Outrageous Ideas) co-located with 18th International Semantic Web Conference (ISWC 2019) (ISWC 2019 Satellites)}, EDITOR = {Su{\'a}rez-Figueroa, Mari Carmen and Cheng, Gong and Gentile, Anna Lisa and Gu{\'e}ret, Christophe and Keet, Maria and Bernstein, Abraham}, PAGES = {237--240}, EID = {62}, SERIES = {CEUR Workshop Proceedings}, VOLUME = {2456}, ADDRESS = {Auckland, New Zealand}, }
Endnote
%0 Conference Proceedings %A Ernst, Patrick %A Terolli, Erisa %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T LongLife: a Platform for Personalized Searchfor Health and Life Sciences : %G eng %U http://hdl.handle.net/21.11116/0000-0005-83A6-B %U http://ceur-ws.org/Vol-2456/paper62.pdf %D 2019 %B 18th Semantic Web Conference %Z date of event: 2019-10-26 - 2019-10-30 %C Auckland, New Zealand %B Proceedings of the ISWC 2019 Satellite Tracks (Posters & Demonstrations, Industry, and Outrageous Ideas) co-located with 18th International Semantic Web Conference (ISWC 2019) %E Suárez-Figueroa, Mari Carmen; Cheng, Gong; Gentile, Anna Lisa; Guéret, Christophe; Keet, Maria; Bernstein, Abraham %P 237 - 240 %Z sequence number: 62 %I ceur-ws.org %B CEUR Workshop Proceedings %N 2456 %@ false
[15]
M. H. Gad-Elrab, D. Stepanova, J. Urbani, and G. Weikum, “Tracy: Tracing Facts over Knowledge Graphs and Text,” in Proceedings of The World Wide Web Conference (WWW 2019), San Francisco, CA, USA, 2019.
Export
BibTeX
@inproceedings{Gad-Elrab_WWW2019, TITLE = {Tracy: {T}racing Facts over Knowledge Graphs and Text}, AUTHOR = {Gad-Elrab, Mohamed Hassan and Stepanova, Daria and Urbani, Jacopo and Weikum, Gerhard}, LANGUAGE = {eng}, ISBN = {978-1-4503-6674-8}, DOI = {10.1145/3308558.3314126}, PUBLISHER = {ACM}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {Proceedings of The World Wide Web Conference (WWW 2019)}, EDITOR = {McAuley, Julian}, PAGES = {3516--3520}, ADDRESS = {San Francisco, CA, USA}, }
Endnote
%0 Conference Proceedings %A Gad-Elrab, Mohamed Hassan %A Stepanova, Daria %A Urbani, Jacopo %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society %T Tracy: Tracing Facts over Knowledge Graphs and Text : %G eng %U http://hdl.handle.net/21.11116/0000-0003-08AA-5 %R 10.1145/3308558.3314126 %D 2019 %B The Web Conference %Z date of event: 2019-05-13 - 2019-05-17 %C San Francisco, CA, USA %B Proceedings of The World Wide Web Conference %E McAuley, Julian %P 3516 - 3520 %I ACM %@ 978-1-4503-6674-8
[16]
M. H. Gad-Elrab, D. Stepanova, J. Urbani, and G. Weikum, “ExFaKT: A Framework for Explaining Facts over Knowledge Graphs and Text ,” in WSDM’19, 12h ACM International Conference on Web Search and Data Mining, Melbourne, Australia, 2019.
Export
BibTeX
@inproceedings{Gad-Elrab_WSDM2019, TITLE = {{ExFaKT}: {A} Framework for Explaining Facts over Knowledge Graphs and Text}, AUTHOR = {Gad-Elrab, Mohamed Hassan and Stepanova, Daria and Urbani, Jacopo and Weikum, Gerhard}, LANGUAGE = {eng}, ISBN = {978-1-4503-5940-5}, DOI = {10.1145/3289600.3290996}, PUBLISHER = {ACM}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, BOOKTITLE = {WSDM'19, 12h ACM International Conference on Web Search and Data Mining}, PAGES = {87--95}, ADDRESS = {Melbourne, Australia}, }
Endnote
%0 Conference Proceedings %A Gad-Elrab, Mohamed Hassan %A Stepanova, Daria %A Urbani, Jacopo %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society %T ExFaKT: A Framework for Explaining Facts over Knowledge Graphs and Text  : %G eng %U http://hdl.handle.net/21.11116/0000-0002-9C44-2 %R 10.1145/3289600.3290996 %D 2019 %B 12h ACM International Conference on Web Search and Data Mining %Z date of event: 2019-02-11 - 2019-02-15 %C Melbourne, Australia %B WSDM'19 %P 87 - 95 %I ACM %@ 978-1-4503-5940-5
[17]
A. Ghazimatin, O. Balalau, R. Saha Roy, and G. Weikum, “PRINCE: Provider-side Interpretability with Counterfactual Explanations in Recommender Systems,” 2019. [Online]. Available: http://arxiv.org/abs/1911.08378. (arXiv: 1911.08378)
Abstract
Interpretable explanations for recommender systems and other machine learning models are crucial to gain user trust. Prior works that have focused on paths connecting users and items in a heterogeneous network have several limitations, such as discovering relationships rather than true explanations, or disregarding other users' privacy. In this work, we take a fresh perspective, and present PRINCE: a provider-side mechanism to produce tangible explanations for end-users, where an explanation is defined to be a set of minimal actions performed by the user that, if removed, changes the recommendation to a different item. Given a recommendation, PRINCE uses a polynomial-time optimal algorithm for finding this minimal set of a user's actions from an exponential search space, based on random walks over dynamic graphs. Experiments on two real-world datasets show that PRINCE provides more compact explanations than intuitive baselines, and insights from a crowdsourced user-study demonstrate the viability of such action-based explanations. We thus posit that PRINCE produces scrutable, actionable, and concise explanations, owing to its use of counterfactual evidence, a user's own actions, and minimal sets, respectively.
Export
BibTeX
@online{Ghazimatin_arXiv1911.08378, TITLE = {{PRINCE}: Provider-side Interpretability with Counterfactual Explanations in Recommender Systems}, AUTHOR = {Ghazimatin, Azin and Balalau, Oana and Saha Roy, Rishiraj and Weikum, Gerhard}, LANGUAGE = {eng}, URL = {http://arxiv.org/abs/1911.08378}, EPRINT = {1911.08378}, EPRINTTYPE = {arXiv}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, ABSTRACT = {Interpretable explanations for recommender systems and other machine learning models are crucial to gain user trust. Prior works that have focused on paths connecting users and items in a heterogeneous network have several limitations, such as discovering relationships rather than true explanations, or disregarding other users' privacy. In this work, we take a fresh perspective, and present PRINCE: a provider-side mechanism to produce tangible explanations for end-users, where an explanation is defined to be a set of minimal actions performed by the user that, if removed, changes the recommendation to a different item. Given a recommendation, PRINCE uses a polynomial-time optimal algorithm for finding this minimal set of a user's actions from an exponential search space, based on random walks over dynamic graphs. Experiments on two real-world datasets show that PRINCE provides more compact explanations than intuitive baselines, and insights from a crowdsourced user-study demonstrate the viability of such action-based explanations. We thus posit that PRINCE produces scrutable, actionable, and concise explanations, owing to its use of counterfactual evidence, a user's own actions, and minimal sets, respectively.}, }
Endnote
%0 Report %A Ghazimatin, Azin %A Balalau, Oana %A Saha Roy, Rishiraj %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T PRINCE: Provider-side Interpretability with Counterfactual Explanations in Recommender Systems : %G eng %U http://hdl.handle.net/21.11116/0000-0005-8415-E %U http://arxiv.org/abs/1911.08378 %D 2019 %X Interpretable explanations for recommender systems and other machine learning models are crucial to gain user trust. Prior works that have focused on paths connecting users and items in a heterogeneous network have several limitations, such as discovering relationships rather than true explanations, or disregarding other users' privacy. In this work, we take a fresh perspective, and present PRINCE: a provider-side mechanism to produce tangible explanations for end-users, where an explanation is defined to be a set of minimal actions performed by the user that, if removed, changes the recommendation to a different item. Given a recommendation, PRINCE uses a polynomial-time optimal algorithm for finding this minimal set of a user's actions from an exponential search space, based on random walks over dynamic graphs. Experiments on two real-world datasets show that PRINCE provides more compact explanations than intuitive baselines, and insights from a crowdsourced user-study demonstrate the viability of such action-based explanations. We thus posit that PRINCE produces scrutable, actionable, and concise explanations, owing to its use of counterfactual evidence, a user's own actions, and minimal sets, respectively. %K Computer Science, Learning, cs.LG,Computer Science, Artificial Intelligence, cs.AI,Statistics, Machine Learning, stat.ML
[18]
A. Ghazimatin, R. Saha Roy, and G. Weikum, “FAIRY: A Framework for Understanding Relationships between Users’ Actions and their Social Feeds,” 2019. [Online]. Available: http://arxiv.org/abs/1908.03109. (arXiv: 1908.03109)
Abstract
Users increasingly rely on social media feeds for consuming daily information. The items in a feed, such as news, questions, songs, etc., usually result from the complex interplay of a user's social contacts, her interests and her actions on the platform. The relationship of the user's own behavior and the received feed is often puzzling, and many users would like to have a clear explanation on why certain items were shown to them. Transparency and explainability are key concerns in the modern world of cognitive overload, filter bubbles, user tracking, and privacy risks. This paper presents FAIRY, a framework that systematically discovers, ranks, and explains relationships between users' actions and items in their social media feeds. We model the user's local neighborhood on the platform as an interaction graph, a form of heterogeneous information network constructed solely from information that is easily accessible to the concerned user. We posit that paths in this interaction graph connecting the user and her feed items can act as pertinent explanations for the user. These paths are scored with a learning-to-rank model that captures relevance and surprisal. User studies on two social platforms demonstrate the practical viability and user benefits of the FAIRY method.
Export
BibTeX
@online{Ghazimatin_arXiv1908.03109, TITLE = {{FAIRY}: A Framework for Understanding Relationships between Users' Actions and their Social Feeds}, AUTHOR = {Ghazimatin, Azin and Saha Roy, Rishiraj and Weikum, Gerhard}, LANGUAGE = {eng}, URL = {http://arxiv.org/abs/1908.03109}, EPRINT = {1908.03109}, EPRINTTYPE = {arXiv}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, ABSTRACT = {Users increasingly rely on social media feeds for consuming daily information. The items in a feed, such as news, questions, songs, etc., usually result from the complex interplay of a user's social contacts, her interests and her actions on the platform. The relationship of the user's own behavior and the received feed is often puzzling, and many users would like to have a clear explanation on why certain items were shown to them. Transparency and explainability are key concerns in the modern world of cognitive overload, filter bubbles, user tracking, and privacy risks. This paper presents FAIRY, a framework that systematically discovers, ranks, and explains relationships between users' actions and items in their social media feeds. We model the user's local neighborhood on the platform as an interaction graph, a form of heterogeneous information network constructed solely from information that is easily accessible to the concerned user. We posit that paths in this interaction graph connecting the user and her feed items can act as pertinent explanations for the user. These paths are scored with a learning-to-rank model that captures relevance and surprisal. User studies on two social platforms demonstrate the practical viability and user benefits of the FAIRY method.}, }
Endnote
%0 Report %A Ghazimatin, Azin %A Saha Roy, Rishiraj %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T FAIRY: A Framework for Understanding Relationships between Users' Actions and their Social Feeds : %G eng %U http://hdl.handle.net/21.11116/0000-0005-83B9-6 %U http://arxiv.org/abs/1908.03109 %D 2019 %X Users increasingly rely on social media feeds for consuming daily information. The items in a feed, such as news, questions, songs, etc., usually result from the complex interplay of a user's social contacts, her interests and her actions on the platform. The relationship of the user's own behavior and the received feed is often puzzling, and many users would like to have a clear explanation on why certain items were shown to them. Transparency and explainability are key concerns in the modern world of cognitive overload, filter bubbles, user tracking, and privacy risks. This paper presents FAIRY, a framework that systematically discovers, ranks, and explains relationships between users' actions and items in their social media feeds. We model the user's local neighborhood on the platform as an interaction graph, a form of heterogeneous information network constructed solely from information that is easily accessible to the concerned user. We posit that paths in this interaction graph connecting the user and her feed items can act as pertinent explanations for the user. These paths are scored with a learning-to-rank model that captures relevance and surprisal. User studies on two social platforms demonstrate the practical viability and user benefits of the FAIRY method. %K cs.SI,Computer Science, Learning, cs.LG,Statistics, Machine Learning, stat.ML,
[19]
A. Ghazimatin, R. Saha Roy, and G. Weikum, “FAIRY: A Framework for Understanding Relationships between Users’ Actions and their Social Feeds,” in WSDM’19, 12h ACM International Conference on Web Search and Data Mining, Melbourne, Australia, 2019.
Export
BibTeX
@inproceedings{Ghazimatin_WSDM2019, TITLE = {{FAIRY}: {A} Framework for Understanding Relationships between Users' Actions and their Social Feeds}, AUTHOR = {Ghazimatin, Azin and Saha Roy, Rishiraj and Weikum, Gerhard}, LANGUAGE = {eng}, ISBN = {978-1-4503-5940-5}, DOI = {10.1145/3289600.3290990}, PUBLISHER = {ACM}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, BOOKTITLE = {WSDM'19, 12h ACM International Conference on Web Search and Data Mining}, PAGES = {240--248}, ADDRESS = {Melbourne, Australia}, }
Endnote
%0 Conference Proceedings %A Ghazimatin, Azin %A Saha Roy, Rishiraj %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T FAIRY: A Framework for Understanding Relationships between Users' Actions and their Social Feeds : %G eng %U http://hdl.handle.net/21.11116/0000-0002-9BD9-B %R 10.1145/3289600.3290990 %D 2019 %B 12h ACM International Conference on Web Search and Data Mining %Z date of event: 2019-02-11 - 2019-02-15 %C Melbourne, Australia %B WSDM'19 %P 240 - 248 %I ACM %@ 978-1-4503-5940-5
[20]
A. Guimarães, O. Balalau, E. Terolli, and G. Weikum, “Analyzing the Traits and Anomalies of Political Discussions on Reddit,” in Proceedings of the Thirteenth International Conference on Web and Social Media (ICWSM 2019), Munich, Germany, 2019.
Export
BibTeX
@inproceedings{Guimaraes_ICWSM2019, TITLE = {Analyzing the Traits and Anomalies of Political Discussions on {R}eddit}, AUTHOR = {Guimar{\~a}es, Anna and Balalau, Oana and Terolli, Erisa and Weikum, Gerhard}, LANGUAGE = {eng}, ISSN = {2334-0770}, PUBLISHER = {AAAI}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {Proceedings of the Thirteenth International Conference on Web and Social Media (ICWSM 2019)}, PAGES = {205--213}, ADDRESS = {Munich, Germany}, }
Endnote
%0 Conference Proceedings %A Guimarães, Anna %A Balalau, Oana %A Terolli, Erisa %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T Analyzing the Traits and Anomalies of Political Discussions on Reddit : %G eng %U http://hdl.handle.net/21.11116/0000-0003-3649-F %D 2019 %B 13th International Conference on Web and Social Media %Z date of event: 2019-06-11 - 2019-06-14 %C Munich, Germany %B Proceedings of the Thirteenth International Conference on Web and Social Media %P 205 - 213 %I AAAI %@ false
[21]
D. Gupta and K. Berberich, “Structured Search in Annotated Document Collections,” in WSDM’19, 12h ACM International Conference on Web Search and Data Mining, Melbourne, Australia, 2019.
Export
BibTeX
@inproceedings{Gupta_WSDM2019Demo, TITLE = {Structured Search in Annotated Document Collections}, AUTHOR = {Gupta, Dhruv and Berberich, Klaus}, LANGUAGE = {eng}, ISBN = {978-1-4503-5940-5}, DOI = {10.1145/3289600.3290618}, PUBLISHER = {ACM}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, BOOKTITLE = {WSDM'19, 12h ACM International Conference on Web Search and Data Mining}, PAGES = {794--797}, ADDRESS = {Melbourne, Australia}, }
Endnote
%0 Conference Proceedings %A Gupta, Dhruv %A Berberich, Klaus %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T Structured Search in Annotated Document Collections : Demo paper %G eng %U http://hdl.handle.net/21.11116/0000-0002-A8D6-F %R 10.1145/3289600.3290618 %D 2019 %B 12h ACM International Conference on Web Search and Data Mining %Z date of event: 2019-02-11 - 2019-02-15 %C Melbourne, Australia %B WSDM'19 %P 794 - 797 %I ACM %@ 978-1-4503-5940-5
[22]
D. Gupta, “Search and Analytics Using Semantic Annotations,” Universität des Saarlandes, Saarbrücken, 2019.
Abstract
Search systems help users locate relevant information in the form of text documents for keyword queries. Using text alone, it is often difficult to satisfy the user's information need. To discern the user's intent behind queries, we turn to semantic annotations (e.g., named entities and temporal expressions) that natural language processing tools can now deliver with great accuracy. This thesis develops methods and an infrastructure that leverage semantic annotations to efficiently and effectively search large document collections. This thesis makes contributions in three areas: indexing, querying, and mining of semantically annotated document collections. First, we describe an indexing infrastructure for semantically annotated document collections. The indexing infrastructure can support knowledge-centric tasks such as information extraction, relationship extraction, question answering, fact spotting and semantic search at scale across millions of documents. Second, we propose methods for exploring large document collections by suggesting semantic aspects for queries. These semantic aspects are generated by considering annotations in the form of temporal expressions, geographic locations, and other named entities. The generated aspects help guide the user to relevant documents without the need to read their contents. Third and finally, we present methods that can generate events, structured tables, and insightful visualizations from semantically annotated document collections.
Export
BibTeX
@phdthesis{GUPTAphd2019, TITLE = {Search and Analytics Using Semantic Annotations}, AUTHOR = {Gupta, Dhruv}, LANGUAGE = {eng}, URL = {urn:nbn:de:bsz:291--ds-300780}, DOI = {10.22028/D291-30078}, SCHOOL = {Universit{\"a}t des Saarlandes}, ADDRESS = {Saarbr{\"u}cken}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, ABSTRACT = {Search systems help users locate relevant information in the form of text documents for keyword queries. Using text alone, it is often difficult to satisfy the user's information need. To discern the user's intent behind queries, we turn to semantic annotations (e.g., named entities and temporal expressions) that natural language processing tools can now deliver with great accuracy. This thesis develops methods and an infrastructure that leverage semantic annotations to efficiently and effectively search large document collections. This thesis makes contributions in three areas: indexing, querying, and mining of semantically annotated document collections. First, we describe an indexing infrastructure for semantically annotated document collections. The indexing infrastructure can support knowledge-centric tasks such as information extraction, relationship extraction, question answering, fact spotting and semantic search at scale across millions of documents. Second, we propose methods for exploring large document collections by suggesting semantic aspects for queries. These semantic aspects are generated by considering annotations in the form of temporal expressions, geographic locations, and other named entities. The generated aspects help guide the user to relevant documents without the need to read their contents. Third and finally, we present methods that can generate events, structured tables, and insightful visualizations from semantically annotated document collections.}, }
Endnote
%0 Thesis %A Gupta, Dhruv %Y Berberich, Klaus %A referee: Weikum, Gerhard %A referee: Bedathur, Srikanta %+ Databases and Information Systems, MPI for Informatics, Max Planck Society International Max Planck Research School, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T Search and Analytics Using Semantic Annotations : %G eng %U http://hdl.handle.net/21.11116/0000-0005-7695-E %R 10.22028/D291-30078 %U urn:nbn:de:bsz:291--ds-300780 %F OTHER: hdl:20.500.11880/28516 %I Universität des Saarlandes %C Saarbrücken %D 2019 %P xxviii, 211 p. %V phd %9 phd %X Search systems help users locate relevant information in the form of text documents for keyword queries. Using text alone, it is often difficult to satisfy the user's information need. To discern the user's intent behind queries, we turn to semantic annotations (e.g., named entities and temporal expressions) that natural language processing tools can now deliver with great accuracy. This thesis develops methods and an infrastructure that leverage semantic annotations to efficiently and effectively search large document collections. This thesis makes contributions in three areas: indexing, querying, and mining of semantically annotated document collections. First, we describe an indexing infrastructure for semantically annotated document collections. The indexing infrastructure can support knowledge-centric tasks such as information extraction, relationship extraction, question answering, fact spotting and semantic search at scale across millions of documents. Second, we propose methods for exploring large document collections by suggesting semantic aspects for queries. These semantic aspects are generated by considering annotations in the form of temporal expressions, geographic locations, and other named entities. The generated aspects help guide the user to relevant documents without the need to read their contents. Third and finally, we present methods that can generate events, structured tables, and insightful visualizations from semantically annotated document collections. %U https://publikationen.sulb.uni-saarland.de/handle/20.500.11880/28516
[23]
D. Gupta, “Search and Analytics Using Semantic Annotations,” ACM SIGIR Forum, vol. 53, no. 2, 2019.
Abstract
Search systems help users locate relevant information in the form of text documents for keyword queries. Using text alone, it is often difficult to satisfy the user's information need. To discern the user's intent behind queries, we turn to semantic annotations (e.g., named entities and temporal expressions) that natural language processing tools can now deliver with great accuracy. This thesis develops methods and an infrastructure that leverage semantic annotations to efficiently and effectively search large document collections. This thesis makes contributions in three areas: indexing, querying, and mining of semantically annotated document collections. First, we describe an indexing infrastructure for semantically annotated document collections. The indexing infrastructure can support knowledge-centric tasks such as information extraction, relationship extraction, question answering, fact spotting and semantic search at scale across millions of documents. Second, we propose methods for exploring large document collections by suggesting semantic aspects for queries. These semantic aspects are generated by considering annotations in the form of temporal expressions, geographic locations, and other named entities. The generated aspects help guide the user to relevant documents without the need to read their contents. Third and finally, we present methods that can generate events, structured tables, and insightful visualizations from semantically annotated document collections.
Export
BibTeX
@article{Gupta_SIGIR19, TITLE = {Search and Analytics Using Semantic Annotations}, AUTHOR = {Gupta, Dhruv}, LANGUAGE = {eng}, PUBLISHER = {ACM}, ADDRESS = {New York, NY}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, ABSTRACT = {Search systems help users locate relevant information in the form of text documents for keyword queries. Using text alone, it is often difficult to satisfy the user's information need. To discern the user's intent behind queries, we turn to semantic annotations (e.g., named entities and temporal expressions) that natural language processing tools can now deliver with great accuracy. This thesis develops methods and an infrastructure that leverage semantic annotations to efficiently and effectively search large document collections. This thesis makes contributions in three areas: indexing, querying, and mining of semantically annotated document collections. First, we describe an indexing infrastructure for semantically annotated document collections. The indexing infrastructure can support knowledge-centric tasks such as information extraction, relationship extraction, question answering, fact spotting and semantic search at scale across millions of documents. Second, we propose methods for exploring large document collections by suggesting semantic aspects for queries. These semantic aspects are generated by considering annotations in the form of temporal expressions, geographic locations, and other named entities. The generated aspects help guide the user to relevant documents without the need to read their contents. Third and finally, we present methods that can generate events, structured tables, and insightful visualizations from semantically annotated document collections.}, JOURNAL = {ACM SIGIR Forum}, VOLUME = {53}, NUMBER = {2}, PAGES = {100--101}, }
Endnote
%0 Journal Article %A Gupta, Dhruv %+ Databases and Information Systems, MPI for Informatics, Max Planck Society International Max Planck Research School, MPI for Informatics, Max Planck Society %T Search and Analytics Using Semantic Annotations : Doctorial Abstract %G eng %U http://hdl.handle.net/21.11116/0000-0005-A1C2-9 %7 2019 %D 2019 %X Search systems help users locate relevant information in the form of text documents for keyword queries. Using text alone, it is often difficult to satisfy the user's information need. To discern the user's intent behind queries, we turn to semantic annotations (e.g., named entities and temporal expressions) that natural language processing tools can now deliver with great accuracy. This thesis develops methods and an infrastructure that leverage semantic annotations to efficiently and effectively search large document collections. This thesis makes contributions in three areas: indexing, querying, and mining of semantically annotated document collections. First, we describe an indexing infrastructure for semantically annotated document collections. The indexing infrastructure can support knowledge-centric tasks such as information extraction, relationship extraction, question answering, fact spotting and semantic search at scale across millions of documents. Second, we propose methods for exploring large document collections by suggesting semantic aspects for queries. These semantic aspects are generated by considering annotations in the form of temporal expressions, geographic locations, and other named entities. The generated aspects help guide the user to relevant documents without the need to read their contents. Third and finally, we present methods that can generate events, structured tables, and insightful visualizations from semantically annotated document collections. %J ACM SIGIR Forum %V 53 %N 2 %& 100 %P 100 - 101 %I ACM %C New York, NY %U http://sigir.org/wp-content/uploads/2019/december/p100.pdf
[24]
D. Gupta and K. Berberich, “JIGSAW: Structuring Text into Tables,” in ICTIR’19, ACM SIGIR International Conference on Theory of Information Retrieval, Santa Clara, CA, USA, 2019.
Export
BibTeX
@inproceedings{Gupta_ICTIR2019, TITLE = {JIGSAW: {S}tructuring Text into Tables}, AUTHOR = {Gupta, Dhruv and Berberich, Klaus}, LANGUAGE = {eng}, ISBN = {978-1-4503-6881-0}, DOI = {10.1145/3341981.3344228}, PUBLISHER = {ACM}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {ICTIR'19, ACM SIGIR International Conference on Theory of Information Retrieval}, EDITOR = {Fang, Yi and Zhang, Yi}, PAGES = {237--244}, ADDRESS = {Santa Clara, CA, USA}, }
Endnote
%0 Conference Proceedings %A Gupta, Dhruv %A Berberich, Klaus %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T JIGSAW: Structuring Text into Tables : %G eng %U http://hdl.handle.net/21.11116/0000-0005-8479-E %R 10.1145/3341981.3344228 %D 2019 %B ACM SIGIR International Conference on Theory of Information Retrieval %Z date of event: 2019-10-02 - 2019-10-05 %C Santa Clara, CA, USA %B ICTIR'19 %E Fang, Yi; Zhang, Yi %P 237 - 244 %I ACM %@ 978-1-4503-6881-0
[25]
D. Gupta, K. Berberich, J. Strötgen, and D. Zeinalipour-Yazti, “Generating Semantic Aspects for Queries,” in The Semantic Web (ESWC 2019), Portorož, Slovenia, 2019.
Export
BibTeX
@inproceedings{GuptaESWC2019, TITLE = {Generating Semantic Aspects for Queries}, AUTHOR = {Gupta, Dhruv and Berberich, Klaus and Str{\"o}tgen, Jannik and Zeinalipour-Yazti, Demetrios}, LANGUAGE = {eng}, ISBN = {978-3-030-21347-3}, DOI = {10.1007/978-3-030-21348-0_11}, PUBLISHER = {Springer}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, BOOKTITLE = {The Semantic Web (ESWC 2019)}, EDITOR = {Hitzler, Pascal and Fern{\'a}ndez, Miriam and Janowicz, Krzysztof and Zaveri, Amrapali and Gray, Alasdair J. G. and Lopez, Vanessa and Haller, Armin and Hammar, Karl}, PAGES = {162--178}, SERIES = {Lecture Notes in Computer Science}, VOLUME = {11503}, ADDRESS = {Portoro{\v z}, Slovenia}, }
Endnote
%0 Conference Proceedings %A Gupta, Dhruv %A Berberich, Klaus %A Strötgen, Jannik %A Zeinalipour-Yazti, Demetrios %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations %T Generating Semantic Aspects for Queries : %G eng %U http://hdl.handle.net/21.11116/0000-0003-FF5F-5 %R 10.1007/978-3-030-21348-0_11 %D 2019 %B 16th Extended Semantic Web Conference %Z date of event: 2019-06-02 - 2019-06-06 %C Portorož, Slovenia %B The Semantic Web %E Hitzler, Pascal; Fernández, Miriam; Janowicz, Krzysztof; Zaveri, Amrapali; Gray, Alasdair J. G.; Lopez, Vanessa; Haller, Armin; Hammar, Karl %P 162 - 178 %I Springer %@ 978-3-030-21347-3 %B Lecture Notes in Computer Science %N 11503
[26]
D. Gupta and K. Berberich, “Efficient Retrieval of Knowledge Graph Fact Evidences,” in The Semantic Web: ESWC 2019 Satellite Events, Portorož, Slovenia, 2019.
Export
BibTeX
@inproceedings{GuptaESWC2019a, TITLE = {Efficient Retrieval of Knowledge Graph Fact Evidences}, AUTHOR = {Gupta, Dhruv and Berberich, Klaus}, LANGUAGE = {eng}, ISBN = {978-3-030-32326-4}, DOI = {10.1007/978-3-030-32327-1_18}, PUBLISHER = {Springer}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, BOOKTITLE = {The Semantic Web: ESWC 2019 Satellite Events}, EDITOR = {Hitzler, Pascal and Kirrane, Sabrina and Hartig, Olaf and de Boer, Victor and Vidal, Maria-Esther and Maleshova, Maria and Schlobach, Stefan and Hammar, Karl and Lasierra, Nelia and Stadtm{\"u}ller, Steffen and Hose, Katja and Verborgh, Ruben}, PAGES = {90--94}, SERIES = {Lecture Notes in Computer Science}, VOLUME = {11762}, ADDRESS = {Portoro{\v z}, Slovenia}, }
Endnote
%0 Conference Proceedings %A Gupta, Dhruv %A Berberich, Klaus %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T Efficient Retrieval of Knowledge Graph Fact Evidences : %G eng %U http://hdl.handle.net/21.11116/0000-0005-8477-0 %R 10.1007/978-3-030-32327-1_18 %D 2019 %B 16th Extended Semantic Web Conference %Z date of event: 2019-06-02 - 2019-06-06 %C Portorož, Slovenia %B The Semantic Web: ESWC 2019 Satellite Events %E Hitzler, Pascal; Kirrane, Sabrina; Hartig, Olaf; de Boer, Victor; Vidal, Maria-Esther; Maleshova, Maria; Schlobach, Stefan; Hammar, Karl; Lasierra, Nelia; Stadtmüller, Steffen; Hose, Katja; Verborgh, Ruben %P 90 - 94 %I Springer %@ 978-3-030-32326-4 %B Lecture Notes in Computer Science %N 11762
[27]
M. A. Hedderich, A. Yates, D. Klakow, and G. de Melo, “Using Multi-Sense Vector Embeddings for Reverse Dictionaries,” in Proceedings of the 13th International Conference on Computational Semantics - Long Papers (IWCS 2019), Gothenburg, Sweden, 2019.
Export
BibTeX
@inproceedings{Hedderich_IWCS2019, TITLE = {Using Multi-Sense Vector Embeddings for Reverse Dictionaries}, AUTHOR = {Hedderich, Michael A. and Yates, Andrew and Klakow, Dietrich and de Melo, Gerard}, LANGUAGE = {eng}, ISBN = {978-1-950737-19-2}, PUBLISHER = {ACL}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {Proceedings of the 13th International Conference on Computational Semantics -- Long Papers (IWCS 2019)}, EDITOR = {Dobnik, Simon and Chatzikyriakidis, Stergios and Demberg, Vera}, PAGES = {247--258}, ADDRESS = {Gothenburg, Sweden}, }
Endnote
%0 Conference Proceedings %A Hedderich, Michael A. %A Yates, Andrew %A Klakow, Dietrich %A de Melo, Gerard %+ External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations %T Using Multi-Sense Vector Embeddings for Reverse Dictionaries : %G eng %U http://hdl.handle.net/21.11116/0000-0004-02A4-0 %D 2019 %B 13th International Conference on Computational Semantics %Z date of event: 2019-05-23 - 2019-05-27 %C Gothenburg, Sweden %B Proceedings of the 13th International Conference on Computational Semantics - Long Papers %E Dobnik, Simon; Chatzikyriakidis, Stergios; Demberg, Vera %P 247 - 258 %I ACL %@ 978-1-950737-19-2 %U https://www.aclweb.org/anthology/W19-0421
[28]
M. A. Hedderich, A. Yates, D. Klakow, and G. de Melo, “Using Multi-Sense Vector Embeddings for Reverse Dictionaries,” 2019. [Online]. Available: http://arxiv.org/abs/1904.01451. (arXiv: 1904.01451)
Abstract
Popular word embedding methods such as word2vec and GloVe assign a single vector representation to each word, even if a word has multiple distinct meanings. Multi-sense embeddings instead provide different vectors for each sense of a word. However, they typically cannot serve as a drop-in replacement for conventional single-sense embeddings, because the correct sense vector needs to be selected for each word. In this work, we study the effect of multi-sense embeddings on the task of reverse dictionaries. We propose a technique to easily integrate them into an existing neural network architecture using an attention mechanism. Our experiments demonstrate that large improvements can be obtained when employing multi-sense embeddings both in the input sequence as well as for the target representation. An analysis of the sense distributions and of the learned attention is provided as well.
Export
BibTeX
@online{Hedderich_arXiv1904.01451, TITLE = {Using Multi-Sense Vector Embeddings for Reverse Dictionaries}, AUTHOR = {Hedderich, Michael A. and Yates, Andrew and Klakow, Dietrich and de Melo, Gerard}, LANGUAGE = {eng}, URL = {http://arxiv.org/abs/1904.01451}, EPRINT = {1904.01451}, EPRINTTYPE = {arXiv}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, ABSTRACT = {Popular word embedding methods such as word2vec and GloVe assign a single vector representation to each word, even if a word has multiple distinct meanings. Multi-sense embeddings instead provide different vectors for each sense of a word. However, they typically cannot serve as a drop-in replacement for conventional single-sense embeddings, because the correct sense vector needs to be selected for each word. In this work, we study the effect of multi-sense embeddings on the task of reverse dictionaries. We propose a technique to easily integrate them into an existing neural network architecture using an attention mechanism. Our experiments demonstrate that large improvements can be obtained when employing multi-sense embeddings both in the input sequence as well as for the target representation. An analysis of the sense distributions and of the learned attention is provided as well.}, }
Endnote
%0 Report %A Hedderich, Michael A. %A Yates, Andrew %A Klakow, Dietrich %A de Melo, Gerard %+ External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations %T Using Multi-Sense Vector Embeddings for Reverse Dictionaries : %G eng %U http://hdl.handle.net/21.11116/0000-0004-02B4-E %U http://arxiv.org/abs/1904.01451 %D 2019 %X Popular word embedding methods such as word2vec and GloVe assign a single vector representation to each word, even if a word has multiple distinct meanings. Multi-sense embeddings instead provide different vectors for each sense of a word. However, they typically cannot serve as a drop-in replacement for conventional single-sense embeddings, because the correct sense vector needs to be selected for each word. In this work, we study the effect of multi-sense embeddings on the task of reverse dictionaries. We propose a technique to easily integrate them into an existing neural network architecture using an attention mechanism. Our experiments demonstrate that large improvements can be obtained when employing multi-sense embeddings both in the input sequence as well as for the target representation. An analysis of the sense distributions and of the learned attention is provided as well. %K Computer Science, Computation and Language, cs.CL,Computer Science, Learning, cs.LG
[29]
V. T. Ho, Y. Ibrahim, K. Pal, K. Berberich, and G. Weikum, “Qsearch: Answering Quantity Queries from Text,” in The Semantic Web -- ISWC 2019, Auckland, New Zealand, 2019.
Export
BibTeX
@inproceedings{Ho_ISWC2019, TITLE = {Qsearch: {A}nswering Quantity Queries from Text}, AUTHOR = {Ho, Vinh Thinh and Ibrahim, Yusra and Pal, Koninika and Berberich, Klaus and Weikum, Gerhard}, LANGUAGE = {eng}, ISSN = {0302-9743}, ISBN = {978-3-030-30792-9}, DOI = {10.1007/978-3-030-30793-6_14}, PUBLISHER = {Springer}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, BOOKTITLE = {The Semantic Web -- ISWC 2019}, DEBUG = {author: Gandon, Fabien}, EDITOR = {Ghidini, Chiara and Hartig, Olaf and Maleshkova, Maria and Sv{\'a}tek, Vojt{\u e}ch and Cruz, Isabel and Hogan, Aidan and Song, Jie and Lefran{\c c}ois, Maxime}, PAGES = {237--257}, EID = {62}, SERIES = {Lecture Notes in Computer Science}, VOLUME = {11778}, ADDRESS = {Auckland, New Zealand}, }
Endnote
%0 Conference Proceedings %A Ho, Vinh Thinh %A Ibrahim, Yusra %A Pal, Koninika %A Berberich, Klaus %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T Qsearch: Answering Quantity Queries from Text : %G eng %U http://hdl.handle.net/21.11116/0000-0005-83AB-6 %R 10.1007/978-3-030-30793-6_14 %D 2019 %B 18th Semantic Web Conference %Z date of event: 2019-10-26 - 2019-10-30 %C Auckland, New Zealand %B The Semantic Web -- ISWC 2019 %E Ghidini, Chiara; Hartig, Olaf; Maleshkova, Maria; Svátek, Vojtĕch; Cruz, Isabel; Hogan, Aidan; Song, Jie; Lefrançois, Maxime; Gandon, Fabien %P 237 - 257 %Z sequence number: 62 %I Springer %@ 978-3-030-30792-9 %B Lecture Notes in Computer Science %N 11778 %@ false
[30]
Y. Ibrahim, “Understanding Quantities in Web Tables and Text,” Universität des Saarlandes, Saarbrücken, 2019.
Abstract
There is a wealth of schema-free tables on the web. The text accompanying these tables explains and qualifies the numerical quantities given in the tables. Despite this ubiquity of tabular data, there is little research that harnesses this wealth of data by semantically understanding the information that is conveyed rather ambiguously in these tables. This information can be disambiguated only by the help of the accompanying text. In the process of understanding quantity mentions in tables and text, we are faced with the following challenges; First, there is no comprehensive knowledge base for anchoring quantity mentions. Second, tables are created ad-hoc without a standard schema and with ambiguous header names; also table cells usually contain abbreviations. Third, quantities can be written in multiple forms and units of measures. Fourth, the text usually refers to the quantities in tables using aggregation, approximation, and different scales. In this thesis, we target these challenges through the following contributions: - We present the Quantity Knowledge Base (QKB), a knowledge base for representing Quantity mentions. We construct the QKB by importing information from Freebase, Wikipedia, and other online sources. - We propose Equity: a system for automatically canonicalizing header names and cell values onto concepts, classes, entities, and uniquely represented quantities registered in a knowledge base. We devise a probabilistic graphical model that captures coherence dependencies between cells in tables and candidate items in the space of concepts, entities, and quantities. Then, we cast the inference problem into an efficient algorithm based on random walks over weighted graphs. baselines. - We introduce the quantity alignment problem: computing bidirectional links between textual mentions of quantities and the corresponding table cells. We propose BriQ: a system for computing such alignments. BriQ copes with the specific challenges of approximate quantities, aggregated quantities, and calculated quantities. - We design ExQuisiTe: a web application that identifies mentions of quantities in text and tables, aligns quantity mentions in the text with related quantity mentions in tables, and generates salient suggestions for extractive text summarization systems.
Export
BibTeX
@phdthesis{yusraphd2019, TITLE = {Understanding Quantities in Web Tables and Text}, AUTHOR = {Ibrahim, Yusra}, LANGUAGE = {eng}, DOI = {10.22028/D291-29657}, SCHOOL = {Universit{\"a}t des Saarlandes}, ADDRESS = {Saarbr{\"u}cken}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, ABSTRACT = {There is a wealth of schema-free tables on the web. The text accompanying these tables explains and qualifies the numerical quantities given in the tables. Despite this ubiquity of tabular data, there is little research that harnesses this wealth of data by semantically understanding the information that is conveyed rather ambiguously in these tables. This information can be disambiguated only by the help of the accompanying text. In the process of understanding quantity mentions in tables and text, we are faced with the following challenges; First, there is no comprehensive knowledge base for anchoring quantity mentions. Second, tables are created ad-hoc without a standard schema and with ambiguous header names; also table cells usually contain abbreviations. Third, quantities can be written in multiple forms and units of measures. Fourth, the text usually refers to the quantities in tables using aggregation, approximation, and different scales. In this thesis, we target these challenges through the following contributions: -- We present the Quantity Knowledge Base (QKB), a knowledge base for representing Quantity mentions. We construct the QKB by importing information from Freebase, Wikipedia, and other online sources. -- We propose Equity: a system for automatically canonicalizing header names and cell values onto concepts, classes, entities, and uniquely represented quantities registered in a knowledge base. We devise a probabilistic graphical model that captures coherence dependencies between cells in tables and candidate items in the space of concepts, entities, and quantities. Then, we cast the inference problem into an efficient algorithm based on random walks over weighted graphs. baselines. -- We introduce the quantity alignment problem: computing bidirectional links between textual mentions of quantities and the corresponding table cells. We propose BriQ: a system for computing such alignments. BriQ copes with the specific challenges of approximate quantities, aggregated quantities, and calculated quantities. -- We design ExQuisiTe: a web application that identifies mentions of quantities in text and tables, aligns quantity mentions in the text with related quantity mentions in tables, and generates salient suggestions for extractive text summarization systems.}, }
Endnote
%0 Thesis %A Ibrahim, Yusra %Y Weikum, Gerhard %A referee: Riedewald, Mirek %A referee: Berberich, Klaus %+ Databases and Information Systems, MPI for Informatics, Max Planck Society International Max Planck Research School, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Algorithms and Complexity, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T Understanding Quantities in Web Tables and Text : %G eng %U http://hdl.handle.net/21.11116/0000-0005-4384-A %R 10.22028/D291-29657 %I Universität des Saarlandes %C Saarbrücken %D 2019 %P 116 p. %V phd %9 phd %X There is a wealth of schema-free tables on the web. The text accompanying these tables explains and qualifies the numerical quantities given in the tables. Despite this ubiquity of tabular data, there is little research that harnesses this wealth of data by semantically understanding the information that is conveyed rather ambiguously in these tables. This information can be disambiguated only by the help of the accompanying text. In the process of understanding quantity mentions in tables and text, we are faced with the following challenges; First, there is no comprehensive knowledge base for anchoring quantity mentions. Second, tables are created ad-hoc without a standard schema and with ambiguous header names; also table cells usually contain abbreviations. Third, quantities can be written in multiple forms and units of measures. Fourth, the text usually refers to the quantities in tables using aggregation, approximation, and different scales. In this thesis, we target these challenges through the following contributions: - We present the Quantity Knowledge Base (QKB), a knowledge base for representing Quantity mentions. We construct the QKB by importing information from Freebase, Wikipedia, and other online sources. - We propose Equity: a system for automatically canonicalizing header names and cell values onto concepts, classes, entities, and uniquely represented quantities registered in a knowledge base. We devise a probabilistic graphical model that captures coherence dependencies between cells in tables and candidate items in the space of concepts, entities, and quantities. Then, we cast the inference problem into an efficient algorithm based on random walks over weighted graphs. baselines. - We introduce the quantity alignment problem: computing bidirectional links between textual mentions of quantities and the corresponding table cells. We propose BriQ: a system for computing such alignments. BriQ copes with the specific challenges of approximate quantities, aggregated quantities, and calculated quantities. - We design ExQuisiTe: a web application that identifies mentions of quantities in text and tables, aligns quantity mentions in the text with related quantity mentions in tables, and generates salient suggestions for extractive text summarization systems. %U https://publikationen.sulb.uni-saarland.de/handle/20.500.11880/28300
[31]
Y. Ibrahim and G. Weikum, “ExQuisiTe: Explaining Quantities in Text,” in Proceedings of the World Wide Web Conference (WWW 2019), San Francisco, CA, USA, 2019.
Export
BibTeX
@inproceedings{Ibrahim_WWW2019, TITLE = {{ExQuisiTe}: {E}xplaining Quantities in Text}, AUTHOR = {Ibrahim, Yusra and Weikum, Gerhard}, LANGUAGE = {eng}, ISBN = {978-1-4503-6674-8}, DOI = {10.1145/3308558.3314134}, PUBLISHER = {ACM}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {Proceedings of the World Wide Web Conference (WWW 2019)}, EDITOR = {McAuley, Julian}, PAGES = {3541--3544}, ADDRESS = {San Francisco, CA, USA}, }
Endnote
%0 Conference Proceedings %A Ibrahim, Yusra %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T ExQuisiTe: Explaining Quantities in Text : %G eng %U http://hdl.handle.net/21.11116/0000-0003-01B3-1 %R 10.1145/3308558.3314134 %D 2019 %B The Web Conference %Z date of event: 2019-05-13 - 2019-05-17 %C San Francisco, CA, USA %B Proceedings of the World Wide Web Conference %E McAuley, Julian %P 3541 - 3544 %I ACM %@ 978-1-4503-6674-8
[32]
Y. Ibrahim, M. Riedewald, G. Weikum, and D. Zeinalipour-Yazti, “Bridging Quantities in Tables and Text,” in ICDE 2019, 35th IEEE International Conference on Data Engineering, Macau, China, 2019.
Export
BibTeX
@inproceedings{Ibrahim_ICDE2019, TITLE = {Bridging Quantities in Tables and Text}, AUTHOR = {Ibrahim, Yusra and Riedewald, Mirek and Weikum, Gerhard and Zeinalipour-Yazti, Demetrios}, LANGUAGE = {eng}, ISBN = {978-1-5386-7474-1}, DOI = {10.1109/ICDE.2019.00094}, PUBLISHER = {IEEE}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {ICDE 2019, 35th IEEE International Conference on Data Engineering}, PAGES = {1010--1021}, ADDRESS = {Macau, China}, }
Endnote
%0 Conference Proceedings %A Ibrahim, Yusra %A Riedewald, Mirek %A Weikum, Gerhard %A Zeinalipour-Yazti, Demetrios %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Algorithms and Complexity, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T Bridging Quantities in Tables and Text : %G eng %U http://hdl.handle.net/21.11116/0000-0003-01AB-B %R 10.1109/ICDE.2019.00094 %D 2019 %B 35th IEEE International Conference on Data Engineering %Z date of event: 2019-04-08 - 2019-04-12 %C Macau, China %B ICDE 2019 %P 1010 - 1021 %I IEEE %@ 978-1-5386-7474-1
[33]
Z. Jia, A. Abujabal, R. Saha Roy, J. Strötgen, and G. Weikum, “TEQUILA: Temporal Question Answering over Knowledge Bases,” 2019. [Online]. Available: http://arxiv.org/abs/1908.03650. (arXiv: 1908.03650)
Abstract
Question answering over knowledge bases (KB-QA) poses challenges in handling complex questions that need to be decomposed into sub-questions. An important case, addressed here, is that of temporal questions, where cues for temporal relations need to be discovered and handled. We present TEQUILA, an enabler method for temporal QA that can run on top of any KB-QA engine. TEQUILA has four stages. It detects if a question has temporal intent. It decomposes and rewrites the question into non-temporal sub-questions and temporal constraints. Answers to sub-questions are then retrieved from the underlying KB-QA engine. Finally, TEQUILA uses constraint reasoning on temporal intervals to compute final answers to the full question. Comparisons against state-of-the-art baselines show the viability of our method.
Export
BibTeX
@online{Jia_arXiv1908.03650, TITLE = {{TEQUILA}: Temporal Question Answering over Knowledge Bases}, AUTHOR = {Jia, Zhen and Abujabal, Abdalghani and Saha Roy, Rishiraj and Str{\"o}tgen, Jannik and Weikum, Gerhard}, LANGUAGE = {eng}, URL = {http://arxiv.org/abs/1908.03650}, EPRINT = {1908.03650}, EPRINTTYPE = {arXiv}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, ABSTRACT = {Question answering over knowledge bases (KB-QA) poses challenges in handling complex questions that need to be decomposed into sub-questions. An important case, addressed here, is that of temporal questions, where cues for temporal relations need to be discovered and handled. We present TEQUILA, an enabler method for temporal QA that can run on top of any KB-QA engine. TEQUILA has four stages. It detects if a question has temporal intent. It decomposes and rewrites the question into non-temporal sub-questions and temporal constraints. Answers to sub-questions are then retrieved from the underlying KB-QA engine. Finally, TEQUILA uses constraint reasoning on temporal intervals to compute final answers to the full question. Comparisons against state-of-the-art baselines show the viability of our method.}, }
Endnote
%0 Report %A Jia, Zhen %A Abujabal, Abdalghani %A Saha Roy, Rishiraj %A Strötgen, Jannik %A Weikum, Gerhard %+ External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society %T TEQUILA: Temporal Question Answering over Knowledge Bases : %G eng %U http://hdl.handle.net/21.11116/0000-0005-83BE-1 %U http://arxiv.org/abs/1908.03650 %D 2019 %X Question answering over knowledge bases (KB-QA) poses challenges in handling complex questions that need to be decomposed into sub-questions. An important case, addressed here, is that of temporal questions, where cues for temporal relations need to be discovered and handled. We present TEQUILA, an enabler method for temporal QA that can run on top of any KB-QA engine. TEQUILA has four stages. It detects if a question has temporal intent. It decomposes and rewrites the question into non-temporal sub-questions and temporal constraints. Answers to sub-questions are then retrieved from the underlying KB-QA engine. Finally, TEQUILA uses constraint reasoning on temporal intervals to compute final answers to the full question. Comparisons against state-of-the-art baselines show the viability of our method. %K Computer Science, Information Retrieval, cs.IR,Computer Science, Computation and Language, cs.CL
[34]
M. Kaiser, R. Saha Roy, and G. Weikum, “CROWN: Conversational Passage Ranking by Reasoning over Word Networks,” 2019. [Online]. Available: http://arxiv.org/abs/1911.02850. (arXiv: 1911.02850)
Abstract
Information needs around a topic cannot be satisfied in a single turn; users typically ask follow-up questions referring to the same theme and a system must be capable of understanding the conversational context of a request to retrieve correct answers. In this paper, we present our submission to the TREC Conversational Assistance Track 2019, in which such a conversational setting is explored. We propose a simple unsupervised method for conversational passage ranking by formulating the passage score for a query as a combination of similarity and coherence. To be specific, passages are preferred that contain words semantically similar to the words used in the question, and where such words appear close by. We built a word-proximity network (WPN) from a large corpus, where words are nodes and there is an edge between two nodes if they co-occur in the same passages in a statistically significant way, within a context window. Our approach, named CROWN, improved nDCG scores over a provided Indri baseline on the CAsT training data. On the evaluation data for CAsT, our best run submission achieved above-average performance with respect to AP@5 and nDCG@1000.
Export
BibTeX
@online{Kaiser_arXiv1911.02850, TITLE = {{CROWN}: Conversational Passage Ranking by Reasoning over Word Networks}, AUTHOR = {Kaiser, Magdalena and Saha Roy, Rishiraj and Weikum, Gerhard}, LANGUAGE = {eng}, URL = {http://arxiv.org/abs/1911.02850}, EPRINT = {1911.02850}, EPRINTTYPE = {arXiv}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, ABSTRACT = {Information needs around a topic cannot be satisfied in a single turn; users typically ask follow-up questions referring to the same theme and a system must be capable of understanding the conversational context of a request to retrieve correct answers. In this paper, we present our submission to the TREC Conversational Assistance Track 2019, in which such a conversational setting is explored. We propose a simple unsupervised method for conversational passage ranking by formulating the passage score for a query as a combination of similarity and coherence. To be specific, passages are preferred that contain words semantically similar to the words used in the question, and where such words appear close by. We built a word-proximity network (WPN) from a large corpus, where words are nodes and there is an edge between two nodes if they co-occur in the same passages in a statistically significant way, within a context window. Our approach, named CROWN, improved nDCG scores over a provided Indri baseline on the CAsT training data. On the evaluation data for CAsT, our best run submission achieved above-average performance with respect to AP@5 and nDCG@1000.}, }
Endnote
%0 Report %A Kaiser, Magdalena %A Saha Roy, Rishiraj %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T CROWN: Conversational Passage Ranking by Reasoning over Word Networks : %G eng %U http://hdl.handle.net/21.11116/0000-0005-83ED-C %U http://arxiv.org/abs/1911.02850 %D 2019 %X Information needs around a topic cannot be satisfied in a single turn; users typically ask follow-up questions referring to the same theme and a system must be capable of understanding the conversational context of a request to retrieve correct answers. In this paper, we present our submission to the TREC Conversational Assistance Track 2019, in which such a conversational setting is explored. We propose a simple unsupervised method for conversational passage ranking by formulating the passage score for a query as a combination of similarity and coherence. To be specific, passages are preferred that contain words semantically similar to the words used in the question, and where such words appear close by. We built a word-proximity network (WPN) from a large corpus, where words are nodes and there is an edge between two nodes if they co-occur in the same passages in a statistically significant way, within a context window. Our approach, named CROWN, improved nDCG scores over a provided Indri baseline on the CAsT training data. On the evaluation data for CAsT, our best run submission achieved above-average performance with respect to AP@5 and nDCG@1000. %K Computer Science, Information Retrieval, cs.IR,Computer Science, Computation and Language, cs.CL
[35]
D. Kaltenpoth and J. Vreeken, “We Are Not Your Real Parents: Telling Causal from Confounded using MDL,” 2019. [Online]. Available: http://arxiv.org/abs/1901.06950. (arXiv: 1901.06950)
Abstract
Given data over variables $(X_1,...,X_m, Y)$ we consider the problem of finding out whether $X$ jointly causes $Y$ or whether they are all confounded by an unobserved latent variable $Z$. To do so, we take an information-theoretic approach based on Kolmogorov complexity. In a nutshell, we follow the postulate that first encoding the true cause, and then the effects given that cause, results in a shorter description than any other encoding of the observed variables. The ideal score is not computable, and hence we have to approximate it. We propose to do so using the Minimum Description Length (MDL) principle. We compare the MDL scores under the models where $X$ causes $Y$ and where there exists a latent variables $Z$ confounding both $X$ and $Y$ and show our scores are consistent. To find potential confounders we propose using latent factor modeling, in particular, probabilistic PCA (PPCA). Empirical evaluation on both synthetic and real-world data shows that our method, CoCa, performs very well -- even when the true generating process of the data is far from the assumptions made by the models we use. Moreover, it is robust as its accuracy goes hand in hand with its confidence.
Export
BibTeX
@online{Kaltenpoth_arXiv1901.06950, TITLE = {We Are Not Your Real Parents: Telling Causal from Confounded using {MDL}}, AUTHOR = {Kaltenpoth, David and Vreeken, Jilles}, LANGUAGE = {eng}, URL = {http://arxiv.org/abs/1901.06950}, EPRINT = {1901.06950}, EPRINTTYPE = {arXiv}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, ABSTRACT = {Given data over variables $(X_1,...,X_m, Y)$ we consider the problem of finding out whether $X$ jointly causes $Y$ or whether they are all confounded by an unobserved latent variable $Z$. To do so, we take an information-theoretic approach based on Kolmogorov complexity. In a nutshell, we follow the postulate that first encoding the true cause, and then the effects given that cause, results in a shorter description than any other encoding of the observed variables. The ideal score is not computable, and hence we have to approximate it. We propose to do so using the Minimum Description Length (MDL) principle. We compare the MDL scores under the models where $X$ causes $Y$ and where there exists a latent variables $Z$ confounding both $X$ and $Y$ and show our scores are consistent. To find potential confounders we propose using latent factor modeling, in particular, probabilistic PCA (PPCA). Empirical evaluation on both synthetic and real-world data shows that our method, CoCa, performs very well -- even when the true generating process of the data is far from the assumptions made by the models we use. Moreover, it is robust as its accuracy goes hand in hand with its confidence.}, }
Endnote
%0 Report %A Kaltenpoth, David %A Vreeken, Jilles %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T We Are Not Your Real Parents: Telling Causal from Confounded using MDL : %G eng %U http://hdl.handle.net/21.11116/0000-0003-FFEE-3 %U http://arxiv.org/abs/1901.06950 %D 2019 %X Given data over variables $(X_1,...,X_m, Y)$ we consider the problem of finding out whether $X$ jointly causes $Y$ or whether they are all confounded by an unobserved latent variable $Z$. To do so, we take an information-theoretic approach based on Kolmogorov complexity. In a nutshell, we follow the postulate that first encoding the true cause, and then the effects given that cause, results in a shorter description than any other encoding of the observed variables. The ideal score is not computable, and hence we have to approximate it. We propose to do so using the Minimum Description Length (MDL) principle. We compare the MDL scores under the models where $X$ causes $Y$ and where there exists a latent variables $Z$ confounding both $X$ and $Y$ and show our scores are consistent. To find potential confounders we propose using latent factor modeling, in particular, probabilistic PCA (PPCA). Empirical evaluation on both synthetic and real-world data shows that our method, CoCa, performs very well -- even when the true generating process of the data is far from the assumptions made by the models we use. Moreover, it is robust as its accuracy goes hand in hand with its confidence. %K Computer Science, Learning, cs.LG,Statistics, Machine Learning, stat.ML
[36]
D. Kaltenpoth and J. Vreeken, “We Are Not Your Real Parents: Telling Causal from Confounded by MDL,” in Proceedings of the 2019 SIAM International Conference on Data Mining (SDM 2019), Calgary, Canada, 2019.
Export
BibTeX
@inproceedings{Kaltenpoth_SDM2019, TITLE = {We Are Not Your Real Parents: {T}elling Causal from Confounded by {MDL}}, AUTHOR = {Kaltenpoth, David and Vreeken, Jilles}, LANGUAGE = {eng}, ISBN = {978-1-61197-567-3}, DOI = {10.1137/1.9781611975673.23}, PUBLISHER = {SIAM}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {Proceedings of the 2019 SIAM International Conference on Data Mining (SDM 2019)}, EDITOR = {Berger-Wolf, Tanya and Chawla, Nitesh}, PAGES = {199--207}, ADDRESS = {Calgary, Canada}, }
Endnote
%0 Conference Proceedings %A Kaltenpoth, David %A Vreeken, Jilles %+ Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations %T We Are Not Your Real Parents: Telling Causal from Confounded by MDL : %G eng %U http://hdl.handle.net/21.11116/0000-0003-0D37-2 %R 10.1137/1.9781611975673.23 %D 2019 %B SIAM International Conference on Data Mining %Z date of event: 2019-05-02 - 2019-05-04 %C Calgary, Canada %B Proceedings of the 2019 SIAM International Conference on Data Mining %E Berger-Wolf, Tanya; Chawla, Nitesh %P 199 - 207 %I SIAM %@ 978-1-61197-567-3
[37]
S. Karaev, “Matrix factorization over diods and its applications in data mining,” Universität des Saarlandes, Saarbrücken, 2019.
Abstract
Matrix factorizations are an important tool in data mining, and they have been used extensively for finding latent patterns in the data. They often allow to separate structure from noise, as well as to considerably reduce the dimensionality of the input matrix. While classical matrix decomposition methods, such as nonnegative matrix factorization (NMF) and singular value decomposition (SVD), proved to be very useful in data analysis, they are limited by the underlying algebraic structure. NMF, in particular, tends to break patterns into smaller bits, often mixing them with each other. This happens because overlapping patterns interfere with each other, making it harder to tell them apart. In this thesis we study matrix factorization over algebraic structures known as dioids, which are characterized by the lack of additive inverse (“negative numbers”) and the idempotency of addition (a + a = a). Using dioids makes it easier to separate overlapping features, and, in particular, it allows to better deal with the above mentioned pattern breaking problem. We consider different types of dioids, that range from continuous (subtropical and tropical algebras) to discrete (Boolean algebra). Among these, the Boolean algebra is perhaps the most well known, and there exist methods that allow one to obtain high quality Boolean matrix factorizations in terms of the reconstruction error. In this work, however, a different objective function is used – the description length of the data, which enables us to obtain compact and highly interpretable results. The tropical and subtropical algebras, on the other hand, are much less known in the data mining field. While they find applications in areas such as job scheduling and discrete event systems, they are virtually unknown in the context of data analysis. We will use them to obtain idempotent nonnegative factorizations that are similar to NMF, but are better at separating the most prominent features of the data.
Export
BibTeX
@phdthesis{Karaevphd2019, TITLE = {Matrix factorization over diods and its applications in data mining}, AUTHOR = {Karaev, Sanjar}, LANGUAGE = {eng}, DOI = {10.22028/D291-28661}, SCHOOL = {Universit{\"a}t des Saarlandes}, ADDRESS = {Saarbr{\"u}cken}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, ABSTRACT = {Matrix factorizations are an important tool in data mining, and they have been used extensively for finding latent patterns in the data. They often allow to separate structure from noise, as well as to considerably reduce the dimensionality of the input matrix. While classical matrix decomposition methods, such as nonnegative matrix factorization (NMF) and singular value decomposition (SVD), proved to be very useful in data analysis, they are limited by the underlying algebraic structure. NMF, in particular, tends to break patterns into smaller bits, often mixing them with each other. This happens because overlapping patterns interfere with each other, making it harder to tell them apart. In this thesis we study matrix factorization over algebraic structures known as dioids, which are characterized by the lack of additive inverse ({\textquotedblleft}negative numbers{\textquotedblright}) and the idempotency of addition (a + a = a). Using dioids makes it easier to separate overlapping features, and, in particular, it allows to better deal with the above mentioned pattern breaking problem. We consider different types of dioids, that range from continuous (subtropical and tropical algebras) to discrete (Boolean algebra). Among these, the Boolean algebra is perhaps the most well known, and there exist methods that allow one to obtain high quality Boolean matrix factorizations in terms of the reconstruction error. In this work, however, a different objective function is used -- the description length of the data, which enables us to obtain compact and highly interpretable results. The tropical and subtropical algebras, on the other hand, are much less known in the data mining field. While they find applications in areas such as job scheduling and discrete event systems, they are virtually unknown in the context of data analysis. We will use them to obtain idempotent nonnegative factorizations that are similar to NMF, but are better at separating the most prominent features of the data.}, }
Endnote
%0 Thesis %A Karaev, Sanjar %Y Miettinen, Pauli %A referee: Weikum, Gerhard %A referee: van Leeuwen, Matthijs %+ Databases and Information Systems, MPI for Informatics, Max Planck Society International Max Planck Research School, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations %T Matrix factorization over diods and its applications in data mining : %G eng %U http://hdl.handle.net/21.11116/0000-0005-4369-A %R 10.22028/D291-28661 %I Universität des Saarlandes %C Saarbrücken %D 2019 %P 113 p. %V phd %9 phd %X Matrix factorizations are an important tool in data mining, and they have been used extensively for finding latent patterns in the data. They often allow to separate structure from noise, as well as to considerably reduce the dimensionality of the input matrix. While classical matrix decomposition methods, such as nonnegative matrix factorization (NMF) and singular value decomposition (SVD), proved to be very useful in data analysis, they are limited by the underlying algebraic structure. NMF, in particular, tends to break patterns into smaller bits, often mixing them with each other. This happens because overlapping patterns interfere with each other, making it harder to tell them apart. In this thesis we study matrix factorization over algebraic structures known as dioids, which are characterized by the lack of additive inverse (“negative numbers”) and the idempotency of addition (a + a = a). Using dioids makes it easier to separate overlapping features, and, in particular, it allows to better deal with the above mentioned pattern breaking problem. We consider different types of dioids, that range from continuous (subtropical and tropical algebras) to discrete (Boolean algebra). Among these, the Boolean algebra is perhaps the most well known, and there exist methods that allow one to obtain high quality Boolean matrix factorizations in terms of the reconstruction error. In this work, however, a different objective function is used – the description length of the data, which enables us to obtain compact and highly interpretable results. The tropical and subtropical algebras, on the other hand, are much less known in the data mining field. While they find applications in areas such as job scheduling and discrete event systems, they are virtually unknown in the context of data analysis. We will use them to obtain idempotent nonnegative factorizations that are similar to NMF, but are better at separating the most prominent features of the data. %U https://publikationen.sulb.uni-saarland.de/handle/20.500.11880/27903
[38]
S. Karaev and P. Miettinen, “Algorithms for Approximate Subtropical Matrix Factorization,” Data Mining and Knowledge Discovery, vol. 33, no. 2, 2019.
Export
BibTeX
@article{Karaev_DMKD2018, TITLE = {Algorithms for Approximate Subtropical Matrix Factorization}, AUTHOR = {Karaev, Sanjar and Miettinen, Pauli}, LANGUAGE = {eng}, DOI = {10.1007/s10618-018-0599-1}, PUBLISHER = {Springer}, ADDRESS = {New York, NY}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, JOURNAL = {Data Mining and Knowledge Discovery}, VOLUME = {33}, NUMBER = {2}, PAGES = {526--576}, }
Endnote
%0 Journal Article %A Karaev, Sanjar %A Miettinen, Pauli %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T Algorithms for Approximate Subtropical Matrix Factorization : %G eng %U http://hdl.handle.net/21.11116/0000-0002-9FD5-B %R 10.1007/s10618-018-0599-1 %7 2018 %D 2019 %J Data Mining and Knowledge Discovery %O DMKD %V 33 %N 2 %& 526 %P 526 - 576 %I Springer %C New York, NY
[39]
A. Konstantinidis, P. Irakleous, Z. Georgiou, D. Zeinalipour-Yazti, and P. K. Chrysanthis, “IoT Data Prefetching in Indoor Navigation SOAs,” ACM Transactions on Internet Technology, vol. 19, no. 1, 2019.
Export
BibTeX
@article{Konstantinidis:2018:IDP:3283809.3177777, TITLE = {{IoT} Data Prefetching in Indoor Navigation {SOAs}}, AUTHOR = {Konstantinidis, Andreas and Irakleous, Panagiotis and Georgiou, Zacharias and Zeinalipour-Yazti, Demetrios and Chrysanthis, Panos K.}, LANGUAGE = {eng}, ISSN = {1533-5399}, DOI = {10.1145/3177777}, PUBLISHER = {ACM}, ADDRESS = {New York, NY}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, JOURNAL = {ACM Transactions on Internet Technology}, VOLUME = {19}, NUMBER = {1}, EID = {10}, }
Endnote
%0 Journal Article %A Konstantinidis, Andreas %A Irakleous, Panagiotis %A Georgiou, Zacharias %A Zeinalipour-Yazti, Demetrios %A Chrysanthis, Panos K. %+ External Organizations External Organizations External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations %T IoT Data Prefetching in Indoor Navigation SOAs : %G eng %U http://hdl.handle.net/21.11116/0000-0002-CA09-1 %R 10.1145/3177777 %7 2019 %D 2019 %J ACM Transactions on Internet Technology %O TOIT %V 19 %N 1 %Z sequence number: 10 %I ACM %C New York, NY %@ false
[40]
P. Lahoti, K. P. Gummadi, and G. Weikum, “Operationalizing Individual Fairness with Pairwise Fair Representations,” 2019. [Online]. Available: http://arxiv.org/abs/1907.01439. (arXiv: 1907.01439)
Abstract
We revisit the notion of individual fairness proposed by Dwork et al. A central challenge in operationalizing their approach is the difficulty in eliciting a human specification of a similarity metric. In this paper, we propose an operationalization of individual fairness that does not rely on a human specification of a distance metric. Instead, we propose novel approaches to elicit and leverage side-information on equally deserving individuals to counter subordination between social groups. We model this knowledge as a fairness graph, and learn a unified Pairwise Fair Representation(PFR) of the data that captures both data-driven similarity between individuals and the pairwise side-information in fairness graph. We elicit fairness judgments from a variety of sources, including humans judgments for two real-world datasets on recidivism prediction (COMPAS) and violent neighborhood prediction (Crime & Communities). Our experiments show that the PFR model for operationalizing individual fairness is practically viable.
Export
BibTeX
@online{Lahoti_arXiv1907.01439, TITLE = {Operationalizing Individual Fairness with Pairwise Fair Representations}, AUTHOR = {Lahoti, Preethi and Gummadi, Krishna P. and Weikum, Gerhard}, LANGUAGE = {eng}, URL = {http://arxiv.org/abs/1907.01439}, EPRINT = {1907.01439}, EPRINTTYPE = {arXiv}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, ABSTRACT = {We revisit the notion of individual fairness proposed by Dwork et al. A central challenge in operationalizing their approach is the difficulty in eliciting a human specification of a similarity metric. In this paper, we propose an operationalization of individual fairness that does not rely on a human specification of a distance metric. Instead, we propose novel approaches to elicit and leverage side-information on equally deserving individuals to counter subordination between social groups. We model this knowledge as a fairness graph, and learn a unified Pairwise Fair Representation(PFR) of the data that captures both data-driven similarity between individuals and the pairwise side-information in fairness graph. We elicit fairness judgments from a variety of sources, including humans judgments for two real-world datasets on recidivism prediction (COMPAS) and violent neighborhood prediction (Crime & Communities). Our experiments show that the PFR model for operationalizing individual fairness is practically viable.}, }
Endnote
%0 Report %A Lahoti, Preethi %A Gummadi, Krishna P. %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society %T Operationalizing Individual Fairness with Pairwise Fair Representations : %G eng %U http://hdl.handle.net/21.11116/0000-0003-FF17-5 %U http://arxiv.org/abs/1907.01439 %D 2019 %X We revisit the notion of individual fairness proposed by Dwork et al. A central challenge in operationalizing their approach is the difficulty in eliciting a human specification of a similarity metric. In this paper, we propose an operationalization of individual fairness that does not rely on a human specification of a distance metric. Instead, we propose novel approaches to elicit and leverage side-information on equally deserving individuals to counter subordination between social groups. We model this knowledge as a fairness graph, and learn a unified Pairwise Fair Representation(PFR) of the data that captures both data-driven similarity between individuals and the pairwise side-information in fairness graph. We elicit fairness judgments from a variety of sources, including humans judgments for two real-world datasets on recidivism prediction (COMPAS) and violent neighborhood prediction (Crime & Communities). Our experiments show that the PFR model for operationalizing individual fairness is practically viable. %K Computer Science, Learning, cs.LG,Statistics, Machine Learning, stat.ML
[41]
P. Lahoti, K. Gummadi, and G. Weikum, “Operationalizing Individual Fairness with Pairwise Fair Representations,” Proceedings of the VLDB Endowment (Proc. VLDB 2019), vol. 13, no. 4, 2019.
Export
BibTeX
@article{Lahoti2019_PVLDB, TITLE = {Operationalizing Individual Fairness with Pairwise Fair Representations}, AUTHOR = {Lahoti, Preethi and Gummadi, Krishna and Weikum, Gerhard}, LANGUAGE = {eng}, DOI = {10.14778/3372716.3372723}, PUBLISHER = {VLDB Endowment Inc.}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, JOURNAL = {Proceedings of the VLDB Endowment (Proc. VLDB)}, VOLUME = {13}, NUMBER = {4}, PAGES = {506--518}, BOOKTITLE = {Proceedings of the 45h International Conference on Very Large Data Bases (VLDB 2019)}, EDITOR = {Balazinska, Magdalena and Zhou, Xiaofang}, }
Endnote
%0 Journal Article %A Lahoti, Preethi %A Gummadi, Krishna %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society %T Operationalizing Individual Fairness with Pairwise Fair Representations : %G eng %U http://hdl.handle.net/21.11116/0000-0005-8168-4 %R 10.14778/3372716.3372723 %7 2019 %D 2019 %J Proceedings of the VLDB Endowment %O PVLDB %V 13 %N 4 %& 506 %P 506 - 518 %I VLDB Endowment Inc. %B Proceedings of the 45h International Conference on Very Large Data Bases %O VLDB 2019 Los Angeles, CA, USA, 26-30 August 2019
[42]
P. Lahoti, K. Gummadi, and G. Weikum, “iFair: Learning Individually Fair Data Representations for Algorithmic Decision Making,” in ICDE 2019, 35th IEEE International Conference on Data Engineering, Macau, China, 2019.
Export
BibTeX
@inproceedings{Lahoti_ICDE2019, TITLE = {{iFair}: {L}earning Individually Fair Data Representations for Algorithmic Decision Making}, AUTHOR = {Lahoti, Preethi and Gummadi, Krishna and Weikum, Gerhard}, LANGUAGE = {eng}, ISBN = {978-1-5386-7474-1}, DOI = {10.1109/ICDE.2019.00121}, PUBLISHER = {IEEE}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {ICDE 2019, 35th IEEE International Conference on Data Engineering}, PAGES = {1334--1345}, ADDRESS = {Macau, China}, }
Endnote
%0 Conference Proceedings %A Lahoti, Preethi %A Gummadi, Krishna %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society %T iFair: Learning Individually Fair Data Representations for Algorithmic Decision Making : %G eng %U http://hdl.handle.net/21.11116/0000-0003-F395-2 %R 10.1109/ICDE.2019.00121 %D 2019 %B 35th IEEE International Conference on Data Engineering %Z date of event: 2019-04-08 - 2019-04-12 %C Macau, China %B ICDE 2019 %P 1334 - 1345 %I IEEE %@ 978-1-5386-7474-1
[43]
X. Lu, S. Pramanik, R. Saha Roy, A. Abujabal, Y. Wang, and G. Weikum, “Answering Complex Questions by Joining Multi-Document Evidence with Quasi Knowledge Graphs,” in SIGIR’19, 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval, Paris, France, 2019.
Export
BibTeX
@inproceedings{lu19answering, TITLE = {Answering Complex Questions by Joining Multi-Document Evidence with Quasi Knowledge Graphs}, AUTHOR = {Lu, Xiaolu and Pramanik, Soumajit and Saha Roy, Rishiraj and Abujabal, Abdalghani and Wang, Yafang and Weikum, Gerhard}, LANGUAGE = {eng}, ISBN = {978-1-4503-6172-9}, DOI = {10.1145/3331184.3331252}, PUBLISHER = {ACM}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {SIGIR'19, 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval}, EDITOR = {Piwowarski, Benjamin and Chevalier, Max and Gaussier, {\'E}ric}, PAGES = {105--114}, ADDRESS = {Paris, France}, }
Endnote
%0 Conference Proceedings %A Lu, Xiaolu %A Pramanik, Soumajit %A Saha Roy, Rishiraj %A Abujabal, Abdalghani %A Wang, Yafang %A Weikum, Gerhard %+ External Organizations External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society %T Answering Complex Questions by Joining Multi-Document Evidence with Quasi Knowledge Graphs : %G eng %U http://hdl.handle.net/21.11116/0000-0003-7085-8 %R 10.1145/3331184.3331252 %D 2019 %B 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval %Z date of event: 2019-07-21 - 2019-07-25 %C Paris, France %B SIGIR'19 %E Piwowarski, Benjamin; Chevalier, Max; Gaussier, Éric %P 105 - 114 %I ACM %@ 978-1-4503-6172-9
[44]
X. Lu, S. Pramanik, R. Saha Roy, A. Abujabal, Y. Wang, and G. Weikum, “Answering Complex Questions by Joining Multi-Document Evidence with Quasi Knowledge Graphs,” SIGIR 2019, 2019. (arXiv: 1908.00469)
Abstract
Direct answering of questions that involve multiple entities and relations is a challenge for text-based QA. This problem is most pronounced when answers can be found only by joining evidence from multiple documents. Curated knowledge graphs (KGs) may yield good answers, but are limited by their inherent incompleteness and potential staleness. This paper presents QUEST, a method that can answer complex questions directly from textual sources on-the-fly, by computing similarity joins over partial results from different documents. Our method is completely unsupervised, avoiding training-data bottlenecks and being able to cope with rapidly evolving ad hoc topics and formulation style in user questions. QUEST builds a noisy quasi KG with node and edge weights, consisting of dynamically retrieved entity names and relational phrases. It augments this graph with types and semantic alignments, and computes the best answers by an algorithm for Group Steiner Trees. We evaluate QUEST on benchmarks of complex questions, and show that it substantially outperforms state-of-the-art baselines.
Export
BibTeX
@article{Lu_arXiv1908.00469, TITLE = {Answering Complex Questions by Joining Multi-Document Evidence with Quasi Knowledge Graphs}, AUTHOR = {Lu, Xiaolu and Pramanik, Soumajit and Saha Roy, Rishiraj and Abujabal, Abdalghani and Wang, Yafang and Weikum, Gerhard}, LANGUAGE = {eng}, URL = {http://arxiv.org/abs/1908.00469}, EPRINT = {1908.00469}, EPRINTTYPE = {arXiv}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, ABSTRACT = {Direct answering of questions that involve multiple entities and relations is a challenge for text-based QA. This problem is most pronounced when answers can be found only by joining evidence from multiple documents. Curated knowledge graphs (KGs) may yield good answers, but are limited by their inherent incompleteness and potential staleness. This paper presents QUEST, a method that can answer complex questions directly from textual sources on-the-fly, by computing similarity joins over partial results from different documents. Our method is completely unsupervised, avoiding training-data bottlenecks and being able to cope with rapidly evolving ad hoc topics and formulation style in user questions. QUEST builds a noisy quasi KG with node and edge weights, consisting of dynamically retrieved entity names and relational phrases. It augments this graph with types and semantic alignments, and computes the best answers by an algorithm for Group Steiner Trees. We evaluate QUEST on benchmarks of complex questions, and show that it substantially outperforms state-of-the-art baselines.}, JOURNAL = {SIGIR 2019}, }
Endnote
%0 Journal Article %A Lu, Xiaolu %A Pramanik, Soumajit %A Saha Roy, Rishiraj %A Abujabal, Abdalghani %A Wang, Yafang %A Weikum, Gerhard %+ External Organizations External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society %T Answering Complex Questions by Joining Multi-Document Evidence with Quasi Knowledge Graphs : %G eng %U http://hdl.handle.net/21.11116/0000-0005-83B3-C %U http://arxiv.org/abs/1908.00469 %7 2019 %D 2019 %X Direct answering of questions that involve multiple entities and relations is a challenge for text-based QA. This problem is most pronounced when answers can be found only by joining evidence from multiple documents. Curated knowledge graphs (KGs) may yield good answers, but are limited by their inherent incompleteness and potential staleness. This paper presents QUEST, a method that can answer complex questions directly from textual sources on-the-fly, by computing similarity joins over partial results from different documents. Our method is completely unsupervised, avoiding training-data bottlenecks and being able to cope with rapidly evolving ad hoc topics and formulation style in user questions. QUEST builds a noisy quasi KG with node and edge weights, consisting of dynamically retrieved entity names and relational phrases. It augments this graph with types and semantic alignments, and computes the best answers by an algorithm for Group Steiner Trees. We evaluate QUEST on benchmarks of complex questions, and show that it substantially outperforms state-of-the-art baselines. %K Computer Science, Information Retrieval, cs.IR %J SIGIR 2019
[45]
S. MacAvaney, A. Yates, A. Cohan, and N. Goharian, “CEDR: Contextualized Embeddings for Document Ranking,” in SIGIR’19, 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval, Paris, France, 2019.
Export
BibTeX
@inproceedings{MacAvaney_SIGIR2019, TITLE = {{CEDR}: Contextualized Embeddings for Document Ranking}, AUTHOR = {MacAvaney, Sean and Yates, Andrew and Cohan, Arman and Goharian, Nazli}, LANGUAGE = {eng}, ISBN = {9781450361729}, DOI = {10.1145/3331184.3331317}, PUBLISHER = {ACM}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, BOOKTITLE = {SIGIR'19, 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval}, PAGES = {1101--1104}, ADDRESS = {Paris, France}, }
Endnote
%0 Conference Proceedings %A MacAvaney, Sean %A Yates, Andrew %A Cohan, Arman %A Goharian, Nazli %+ External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations %T CEDR: Contextualized Embeddings for Document Ranking : %G eng %U http://hdl.handle.net/21.11116/0000-0004-02D3-B %R 10.1145/3331184.3331317 %D 2019 %B 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval %Z date of event: 2019-07-21 - 2019-07-25 %C Paris, France %B SIGIR'19 %P 1101 - 1104 %I ACM %@ 9781450361729
[46]
S. MacAvaney, A. Yates, A. Cohan, and N. Goharian, “CEDR: Contextualized Embeddings for Document Ranking,” 2019. [Online]. Available: http://arxiv.org/abs/1904.07094. (arXiv: 1904.07094)
Abstract
Although considerable attention has been given to neural ranking architectures recently, far less attention has been paid to the term representations that are used as input to these models. In this work, we investigate how two pretrained contextualized language modes (ELMo and BERT) can be utilized for ad-hoc document ranking. Through experiments on TREC benchmarks, we find that several existing neural ranking architectures can benefit from the additional context provided by contextualized language models. Furthermore, we propose a joint approach that incorporates BERT's classification vector into existing neural models and show that it outperforms state-of-the-art ad-hoc ranking baselines. We call this joint approach CEDR (Contextualized Embeddings for Document Ranking). We also address practical challenges in using these models for ranking, including the maximum input length imposed by BERT and runtime performance impacts of contextualized language models.
Export
BibTeX
@online{MacAvaney_arXiv1904.07094, TITLE = {{CEDR}: Contextualized Embeddings for Document Ranking}, AUTHOR = {MacAvaney, Sean and Yates, Andrew and Cohan, Arman and Goharian, Nazli}, LANGUAGE = {eng}, URL = {http://arxiv.org/abs/1904.07094}, EPRINT = {1904.07094}, EPRINTTYPE = {arXiv}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, ABSTRACT = {Although considerable attention has been given to neural ranking architectures recently, far less attention has been paid to the term representations that are used as input to these models. In this work, we investigate how two pretrained contextualized language modes (ELMo and BERT) can be utilized for ad-hoc document ranking. Through experiments on TREC benchmarks, we find that several existing neural ranking architectures can benefit from the additional context provided by contextualized language models. Furthermore, we propose a joint approach that incorporates BERT's classification vector into existing neural models and show that it outperforms state-of-the-art ad-hoc ranking baselines. We call this joint approach CEDR (Contextualized Embeddings for Document Ranking). We also address practical challenges in using these models for ranking, including the maximum input length imposed by BERT and runtime performance impacts of contextualized language models.}, }
Endnote
%0 Report %A MacAvaney, Sean %A Yates, Andrew %A Cohan, Arman %A Goharian, Nazli %+ External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations %T CEDR: Contextualized Embeddings for Document Ranking : %G eng %U http://hdl.handle.net/21.11116/0000-0004-02C7-9 %U http://arxiv.org/abs/1904.07094 %D 2019 %X Although considerable attention has been given to neural ranking architectures recently, far less attention has been paid to the term representations that are used as input to these models. In this work, we investigate how two pretrained contextualized language modes (ELMo and BERT) can be utilized for ad-hoc document ranking. Through experiments on TREC benchmarks, we find that several existing neural ranking architectures can benefit from the additional context provided by contextualized language models. Furthermore, we propose a joint approach that incorporates BERT's classification vector into existing neural models and show that it outperforms state-of-the-art ad-hoc ranking baselines. We call this joint approach CEDR (Contextualized Embeddings for Document Ranking). We also address practical challenges in using these models for ranking, including the maximum input length imposed by BERT and runtime performance impacts of contextualized language models. %K Computer Science, Information Retrieval, cs.IR,Computer Science, Computation and Language, cs.CL
[47]
S. MacAvaney, A. Yates, A. Cohan, L. Soldaini, K. Hui, N. Goharian, and O. Frieder, “Overcoming Low-Utility Facets for Complex Answer Retrieval,” Information Retrieval Journal, vol. 22, no. 3–4, 2019.
Export
BibTeX
@article{MacAvaney2019, TITLE = {Overcoming Low-Utility Facets for Complex Answer Retrieval}, AUTHOR = {MacAvaney, Sean and Yates, Andrew and Cohan, Arman and Soldaini, Luca and Hui, Kai and Goharian, Nazli and Frieder, Ophir}, LANGUAGE = {eng}, ISSN = {1386-4564}, DOI = {10.1007/s10791-018-9343-0}, PUBLISHER = {Springer}, ADDRESS = {New York, NY}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, JOURNAL = {Information Retrieval Journal}, VOLUME = {22}, NUMBER = {3-4}, PAGES = {395--418}, }
Endnote
%0 Journal Article %A MacAvaney, Sean %A Yates, Andrew %A Cohan, Arman %A Soldaini, Luca %A Hui, Kai %A Goharian, Nazli %A Frieder, Ophir %+ External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations External Organizations External Organizations External Organizations %T Overcoming Low-Utility Facets for Complex Answer Retrieval : %G eng %U http://hdl.handle.net/21.11116/0000-0003-C4A1-9 %R 10.1007/s10791-018-9343-0 %7 2019 %D 2019 %J Information Retrieval Journal %V 22 %N 3-4 %& 395 %P 395 - 418 %I Springer %C New York, NY %@ false
[48]
S. MacAvaney, A. Yates, K. Hui, and O. Frieder, “Content-Based Weak Supervision for Ad-Hoc Re-Ranking,” 2019. [Online]. Available: http://arxiv.org/abs/1707.00189. (arXiv: 1707.00189)
Abstract
One challenge with neural ranking is the need for a large amount of manually-labeled relevance judgments for training. In contrast with prior work, we examine the use of weak supervision sources for training that yield pseudo query-document pairs that already exhibit relevance (e.g., newswire headline-content pairs and encyclopedic heading-paragraph pairs). We also propose filtering techniques to eliminate training samples that are too far out of domain using two techniques: a heuristic-based approach and novel supervised filter that re-purposes a neural ranker. Using several leading neural ranking architectures and multiple weak supervision datasets, we show that these sources of training pairs are effective on their own (outperforming prior weak supervision techniques), and that filtering can further improve performance.
Export
BibTeX
@online{MacAvaney_arXiv1707.00189, TITLE = {Content-Based Weak Supervision for Ad-Hoc Re-Ranking}, AUTHOR = {MacAvaney, Sean and Yates, Andrew and Hui, Kai and Frieder, Ophir}, LANGUAGE = {eng}, URL = {http://arxiv.org/abs/1707.00189}, EPRINT = {1707.00189}, EPRINTTYPE = {arXiv}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, ABSTRACT = {One challenge with neural ranking is the need for a large amount of manually-labeled relevance judgments for training. In contrast with prior work, we examine the use of weak supervision sources for training that yield pseudo query-document pairs that already exhibit relevance (e.g., newswire headline-content pairs and encyclopedic heading-paragraph pairs). We also propose filtering techniques to eliminate training samples that are too far out of domain using two techniques: a heuristic-based approach and novel supervised filter that re-purposes a neural ranker. Using several leading neural ranking architectures and multiple weak supervision datasets, we show that these sources of training pairs are effective on their own (outperforming prior weak supervision techniques), and that filtering can further improve performance.}, }
Endnote
%0 Report %A MacAvaney, Sean %A Yates, Andrew %A Hui, Kai %A Frieder, Ophir %+ External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations %T Content-Based Weak Supervision for Ad-Hoc Re-Ranking : %G eng %U http://hdl.handle.net/21.11116/0000-0005-6B59-0 %U http://arxiv.org/abs/1707.00189 %D 2019 %X One challenge with neural ranking is the need for a large amount of manually-labeled relevance judgments for training. In contrast with prior work, we examine the use of weak supervision sources for training that yield pseudo query-document pairs that already exhibit relevance (e.g., newswire headline-content pairs and encyclopedic heading-paragraph pairs). We also propose filtering techniques to eliminate training samples that are too far out of domain using two techniques: a heuristic-based approach and novel supervised filter that re-purposes a neural ranker. Using several leading neural ranking architectures and multiple weak supervision datasets, we show that these sources of training pairs are effective on their own (outperforming prior weak supervision techniques), and that filtering can further improve performance. %K Computer Science, Information Retrieval, cs.IR,Computer Science, Computation and Language, cs.CL
[49]
S. MacAvaney, A. Yates, K. Hui, and O. Frieder, “Content-Based Weak Supervision for Ad-Hoc Re-Ranking,” in SIGIR’19, 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval, Paris, France, 2019.
Export
BibTeX
@inproceedings{MacAvaney_SIGIR2019b, TITLE = {Content-Based Weak Supervision for Ad-Hoc Re-Ranking}, AUTHOR = {MacAvaney, Sean and Yates, Andrew and Hui, Kai and Frieder, Ophir}, LANGUAGE = {eng}, ISBN = {9781450361729}, DOI = {10.1145/3331184.3331316}, PUBLISHER = {ACM}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, BOOKTITLE = {SIGIR'19, 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval}, PAGES = {993--996}, ADDRESS = {Paris, France}, }
Endnote
%0 Conference Proceedings %A MacAvaney, Sean %A Yates, Andrew %A Hui, Kai %A Frieder, Ophir %+ External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations %T Content-Based Weak Supervision for Ad-Hoc Re-Ranking : %G eng %U http://hdl.handle.net/21.11116/0000-0005-6B55-4 %R 10.1145/3331184.3331316 %D 2019 %B 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval %Z date of event: 2019-07-21 - 2019-07-25 %C Paris, France %B SIGIR'19 %P 993 - 996 %I ACM %@ 9781450361729
[50]
P. Mandros, M. Boley, and J. Vreeken, “Discovering Reliable Dependencies from Data: Hardness and Improved Algorithms,” in Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence (IJCAI 2019), Macao, 2019.
Export
BibTeX
@inproceedings{mandros_IJCAI2019, TITLE = {Discovering Reliable Dependencies from Data: {H}ardness and Improved Algorithms}, AUTHOR = {Mandros, Panagiotis and Boley, Mario and Vreeken, Jilles}, LANGUAGE = {eng}, ISBN = {978-0-9992411-4-1}, DOI = {10.24963/ijcai.2019/864}, PUBLISHER = {IJCAI}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence (IJCAI 2019)}, EDITOR = {Krais, Sarit}, PAGES = {6206--6210}, ADDRESS = {Macao}, }
Endnote
%0 Conference Proceedings %A Mandros, Panagiotis %A Boley, Mario %A Vreeken, Jilles %+ Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations %T Discovering Reliable Dependencies from Data: Hardness and Improved Algorithms : %G eng %U http://hdl.handle.net/21.11116/0000-0005-848A-A %R 10.24963/ijcai.2019/864 %D 2019 %B Twenty-Eighth International Joint Conference on Artificial Intelligence %Z date of event: 2019-08-10 - 2019-08-16 %C Macao %B Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence %E Krais, Sarit %P 6206 - 6210 %I IJCAI %@ 978-0-9992411-4-1 %U https://www.ijcai.org/Proceedings/2019/0864.pdf
[51]
P. Mandros, M. Boley, and J. Vreeken, “Discovering Reliable Correlations in Categorical Data,” 2019. [Online]. Available: http://arxiv.org/abs/1908.11682. (arXiv: 1908.11682)
Abstract
In many scientific tasks we are interested in discovering whether there exist any correlations in our data. This raises many questions, such as how to reliably and interpretably measure correlation between a multivariate set of attributes, how to do so without having to make assumptions on distribution of the data or the type of correlation, and, how to efficiently discover the top-most reliably correlated attribute sets from data. In this paper we answer these questions for discovery tasks in categorical data. In particular, we propose a corrected-for-chance, consistent, and efficient estimator for normalized total correlation, by which we obtain a reliable, naturally interpretable, non-parametric measure for correlation over multivariate sets. For the discovery of the top-k correlated sets, we derive an effective algorithmic framework based on a tight bounding function. This framework offers exact, approximate, and heuristic search. Empirical evaluation shows that already for small sample sizes the estimator leads to low-regret optimization outcomes, while the algorithms are shown to be highly effective for both large and high-dimensional data. Through two case studies we confirm that our discovery framework identifies interesting and meaningful correlations.
Export
BibTeX
@online{Mandros_arXiv1908.11682, TITLE = {Discovering Reliable Correlations in Categorical Data}, AUTHOR = {Mandros, Panagiotis and Boley, Mario and Vreeken, Jilles}, LANGUAGE = {eng}, URL = {http://arxiv.org/abs/1908.11682}, EPRINT = {1908.11682}, EPRINTTYPE = {arXiv}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, ABSTRACT = {In many scientific tasks we are interested in discovering whether there exist any correlations in our data. This raises many questions, such as how to reliably and interpretably measure correlation between a multivariate set of attributes, how to do so without having to make assumptions on distribution of the data or the type of correlation, and, how to efficiently discover the top-most reliably correlated attribute sets from data. In this paper we answer these questions for discovery tasks in categorical data. In particular, we propose a corrected-for-chance, consistent, and efficient estimator for normalized total correlation, by which we obtain a reliable, naturally interpretable, non-parametric measure for correlation over multivariate sets. For the discovery of the top-k correlated sets, we derive an effective algorithmic framework based on a tight bounding function. This framework offers exact, approximate, and heuristic search. Empirical evaluation shows that already for small sample sizes the estimator leads to low-regret optimization outcomes, while the algorithms are shown to be highly effective for both large and high-dimensional data. Through two case studies we confirm that our discovery framework identifies interesting and meaningful correlations.}, }
Endnote
%0 Report %A Mandros, Panagiotis %A Boley, Mario %A Vreeken, Jilles %+ Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations %T Discovering Reliable Correlations in Categorical Data : %G eng %U http://hdl.handle.net/21.11116/0000-0005-8491-1 %U http://arxiv.org/abs/1908.11682 %D 2019 %X In many scientific tasks we are interested in discovering whether there exist any correlations in our data. This raises many questions, such as how to reliably and interpretably measure correlation between a multivariate set of attributes, how to do so without having to make assumptions on distribution of the data or the type of correlation, and, how to efficiently discover the top-most reliably correlated attribute sets from data. In this paper we answer these questions for discovery tasks in categorical data. In particular, we propose a corrected-for-chance, consistent, and efficient estimator for normalized total correlation, by which we obtain a reliable, naturally interpretable, non-parametric measure for correlation over multivariate sets. For the discovery of the top-k correlated sets, we derive an effective algorithmic framework based on a tight bounding function. This framework offers exact, approximate, and heuristic search. Empirical evaluation shows that already for small sample sizes the estimator leads to low-regret optimization outcomes, while the algorithms are shown to be highly effective for both large and high-dimensional data. Through two case studies we confirm that our discovery framework identifies interesting and meaningful correlations. %K Computer Science, Learning, cs.LG,Computer Science, Databases, cs.DB,Computer Science, Information Theory, cs.IT,Mathematics, Information Theory, math.IT,Statistics, Machine Learning, stat.ML
[52]
A. Marx and J. Vreeken, “Testing Conditional Independence on Discrete Data using Stochastic Complexity,” in Proceedings of the 22nd International Conference on Artificial Intelligence and Statistics (AISTATS 2019), Naha, Okinawa, Japan, 2019.
Export
BibTeX
@inproceedings{Marx_AISTATS2019, TITLE = {Testing Conditional Independence on Discrete Data using Stochastic Complexity}, AUTHOR = {Marx, Alexander and Vreeken, Jilles}, LANGUAGE = {eng}, PUBLISHER = {PMLR}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {Proceedings of the 22nd International Conference on Artificial Intelligence and Statistics (AISTATS 2019)}, EDITOR = {Chaudhuri, Kamalika and Sugiyama, Masashi}, PAGES = {496--505}, SERIES = {Proceedings of the Machine Learning Research}, VOLUME = {89}, ADDRESS = {Naha, Okinawa, Japan}, }
Endnote
%0 Conference Proceedings %A Marx, Alexander %A Vreeken, Jilles %+ Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations %T Testing Conditional Independence on Discrete Data using Stochastic Complexity : %G eng %U http://hdl.handle.net/21.11116/0000-0003-0D3C-D %D 2019 %B 22nd International Conference on Artificial Intelligence and Statistics %Z date of event: 2019-04-16 - 2019-04-18 %C Naha, Okinawa, Japan %B Proceedings of the 22nd International Conference on Artificial Intelligence and Statistics %E Chaudhuri, Kamalika; Sugiyama, Masashi %P 496 - 505 %I PMLR %B Proceedings of the Machine Learning Research %N 89 %U http://proceedings.mlr.press/v89/marx19a/marx19a.pdf
[53]
A. Marx and J. Vreeken, “Testing Conditional Independence on Discrete Data using Stochastic Complexity,” 2019. [Online]. Available: http://arxiv.org/abs/1903.04829. (arXiv: 1903.04829)
Abstract
Testing for conditional independence is a core aspect of constraint-based causal discovery. Although commonly used tests are perfect in theory, they often fail to reject independence in practice, especially when conditioning on multiple variables. We focus on discrete data and propose a new test based on the notion of algorithmic independence that we instantiate using stochastic complexity. Amongst others, we show that our proposed test, SCI, is an asymptotically unbiased as well as $L_2$ consistent estimator for conditional mutual information (CMI). Further, we show that SCI can be reformulated to find a sensible threshold for CMI that works well on limited samples. Empirical evaluation shows that SCI has a lower type II error than commonly used tests. As a result, we obtain a higher recall when we use SCI in causal discovery algorithms, without compromising the precision.
Export
BibTeX
@online{Marx_arXiv1903.04829, TITLE = {Testing Conditional Independence on Discrete Data using Stochastic Complexity}, AUTHOR = {Marx, Alexander and Vreeken, Jilles}, LANGUAGE = {eng}, URL = {http://arxiv.org/abs/1903.04829}, EPRINT = {1903.04829}, EPRINTTYPE = {arXiv}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, ABSTRACT = {Testing for conditional independence is a core aspect of constraint-based causal discovery. Although commonly used tests are perfect in theory, they often fail to reject independence in practice, especially when conditioning on multiple variables. We focus on discrete data and propose a new test based on the notion of algorithmic independence that we instantiate using stochastic complexity. Amongst others, we show that our proposed test, SCI, is an asymptotically unbiased as well as $L_2$ consistent estimator for conditional mutual information (CMI). Further, we show that SCI can be reformulated to find a sensible threshold for CMI that works well on limited samples. Empirical evaluation shows that SCI has a lower type II error than commonly used tests. As a result, we obtain a higher recall when we use SCI in causal discovery algorithms, without compromising the precision.}, }
Endnote
%0 Report %A Marx, Alexander %A Vreeken, Jilles %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T Testing Conditional Independence on Discrete Data using Stochastic Complexity : %G eng %U http://hdl.handle.net/21.11116/0000-0004-027A-1 %U http://arxiv.org/abs/1903.04829 %D 2019 %X Testing for conditional independence is a core aspect of constraint-based causal discovery. Although commonly used tests are perfect in theory, they often fail to reject independence in practice, especially when conditioning on multiple variables. We focus on discrete data and propose a new test based on the notion of algorithmic independence that we instantiate using stochastic complexity. Amongst others, we show that our proposed test, SCI, is an asymptotically unbiased as well as $L_2$ consistent estimator for conditional mutual information (CMI). Further, we show that SCI can be reformulated to find a sensible threshold for CMI that works well on limited samples. Empirical evaluation shows that SCI has a lower type II error than commonly used tests. As a result, we obtain a higher recall when we use SCI in causal discovery algorithms, without compromising the precision. %K Statistics, Machine Learning, stat.ML,Computer Science, Learning, cs.LG
[54]
A. Marx and J. Vreeken, “Telling Cause from Effect by Local and Global Regression,” Knowledge and Information Systems, vol. 60, no. 3, 2019.
Export
BibTeX
@article{marx:19:crack, TITLE = {Telling Cause from Effect by Local and Global Regression}, AUTHOR = {Marx, Alexander and Vreeken, Jilles}, LANGUAGE = {eng}, ISSN = {0219-1377}, DOI = {10.1007/s10115-018-1286-7}, PUBLISHER = {Springer}, ADDRESS = {New York, NY}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, JOURNAL = {Knowledge and Information Systems}, VOLUME = {60}, NUMBER = {3}, PAGES = {1277--1305}, }
Endnote
%0 Journal Article %A Marx, Alexander %A Vreeken, Jilles %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T Telling Cause from Effect by Local and Global Regression : %G eng %U http://hdl.handle.net/21.11116/0000-0002-9EAD-A %R 10.1007/s10115-018-1286-7 %7 2018-12-07 %D 2019 %J Knowledge and Information Systems %V 60 %N 3 %& 1277 %P 1277 - 1305 %I Springer %C New York, NY %@ false
[55]
A. Marx and J. Vreeken, “Identifiability of Cause and Effect using Regularized Regression,” in KDD’19, 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, Anchorage, AK, USA, 2019.
Export
BibTeX
@inproceedings{Marx_KDD2019, TITLE = {Identifiability of Cause and Effect using Regularized Regression}, AUTHOR = {Marx, Alexander and Vreeken, Jilles}, LANGUAGE = {eng}, ISBN = {978-1-4503-6201-6}, DOI = {10.1145/3292500.3330854}, PUBLISHER = {ACM}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, BOOKTITLE = {KDD'19, 25th ACM SIGKDD International Conference on Knowledge Discovery \& Data Mining}, PAGES = {852--861}, ADDRESS = {Anchorage, AK, USA}, }
Endnote
%0 Conference Proceedings %A Marx, Alexander %A Vreeken, Jilles %+ Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations %T Identifiability of Cause and Effect using Regularized Regression : %G eng %U http://hdl.handle.net/21.11116/0000-0004-858C-8 %R 10.1145/3292500.3330854 %D 2019 %B 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining %Z date of event: 2019-08-04 - 2019-08-08 %C Anchorage, AK, USA %B KDD'19 %P 852 - 861 %I ACM %@ 978-1-4503-6201-6
[56]
A. Marx and J. Vreeken, “Causal Inference on Multivariate and Mixed-Type Data,” in Machine Learning and Knowledge Discovery in Databases (ECML PKDD 2018), Dublin, Ireland, 2019.
Export
BibTeX
@inproceedings{marx:18:crack, TITLE = {Causal Inference on Multivariate and Mixed-Type Data}, AUTHOR = {Marx, Alexander and Vreeken, Jilles}, LANGUAGE = {eng}, ISBN = {978-3-030-10927-1}, DOI = {10.1007/978-3-030-10928-8_39}, PUBLISHER = {Springer}, YEAR = {2018}, MARGINALMARK = {$\bullet$}, DATE = {2019}, BOOKTITLE = {Machine Learning and Knowledge Discovery in Databases (ECML PKDD 2018)}, EDITOR = {Berlingerio, Michele and Bonchi, Francesco and G{\"a}rtner, Thomas and Hurley, Neil and Ifrim, Georgiana}, PAGES = {655--671}, SERIES = {Lecture Notes in Artificial Intelligence}, VOLUME = {11052}, ADDRESS = {Dublin, Ireland}, }
Endnote
%0 Conference Proceedings %A Marx, Alexander %A Vreeken, Jilles %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T Causal Inference on Multivariate and Mixed-Type Data : %G eng %U http://hdl.handle.net/21.11116/0000-0002-9E86-5 %R 10.1007/978-3-030-10928-8_39 %D 2019 %B European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases %Z date of event: 2018-09-10 - 2018-09-14 %C Dublin, Ireland %B Machine Learning and Knowledge Discovery in Databases %E Berlingerio, Michele; Bonchi, Francesco; Gärtner, Thomas; Hurley, Neil; Ifrim, Georgiana %P 655 - 671 %I Springer %@ 978-3-030-10927-1 %B Lecture Notes in Artificial Intelligence %N 11052
[57]
A. Marx and J. Vreeken, “Approximating Algorithmic Conditional Independence for Discrete Data,” in Proceedings of the First AAAI Spring Symposium Beyond Curve Fitting: Causation, Counterfactuals, and Imagination-based AI, Stanford, CA, USA. (Accepted/in press)
Export
BibTeX
@inproceedings{Marx_AAAISpringSymp2019, TITLE = {Approximating Algorithmic Conditional Independence for Discrete Data}, AUTHOR = {Marx, Alexander and Vreeken, Jilles}, LANGUAGE = {eng}, YEAR = {2019}, PUBLREMARK = {Accepted}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {Proceedings of the First AAAI Spring Symposium Beyond Curve Fitting: Causation, Counterfactuals, and Imagination-based AI}, ADDRESS = {Stanford, CA, USA}, }
Endnote
%0 Conference Proceedings %A Marx, Alexander %A Vreeken, Jilles %+ Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations %T Approximating Algorithmic Conditional Independence for Discrete Data : %G eng %U http://hdl.handle.net/21.11116/0000-0003-0D4C-B %D 2019 %B First AAAI Spring Symposium Beyond Curve Fitting: Causation, Counterfactuals, and Imagination-based AI %Z date of event: 2019-05-25 - 2019-05-27 %C Stanford, CA, USA %B Proceedings of the First AAAI Spring Symposium Beyond Curve Fitting: Causation, Counterfactuals, and Imagination-based AI
[58]
S. Metzler, S. Günnemann, and P. Miettinen, “Stability and Dynamics of Communities on Online Question-Answer Sites,” Social Networks, vol. 58, 2019.
Export
BibTeX
@article{Metzler2019, TITLE = {Stability and Dynamics of Communities on Online Question-Answer Sites}, AUTHOR = {Metzler, Saskia and G{\"u}nnemann, Stephan and Miettinen, Pauli}, LANGUAGE = {eng}, ISSN = {0378-8733}, DOI = {10.1016/j.socnet.2018.12.004}, PUBLISHER = {Elsevier}, ADDRESS = {Amsterdam}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, JOURNAL = {Social Networks}, VOLUME = {58}, PAGES = {50--58}, }
Endnote
%0 Journal Article %A Metzler, Saskia %A Günnemann, Stephan %A Miettinen, Pauli %+ Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society %T Stability and Dynamics of Communities on Online Question-Answer Sites : %G eng %U http://hdl.handle.net/21.11116/0000-0002-BCC1-0 %R 10.1016/j.socnet.2018.12.004 %7 2019 %D 2019 %J Social Networks %V 58 %& 50 %P 50 - 58 %I Elsevier %C Amsterdam %@ false
[59]
S. Metzler and P. Miettinen, “HyGen: Generating Random Graphs with Hyperbolic Communities,” Applied Network Science, vol. 4, 2019.
Export
BibTeX
@article{Metzler_Miettienen19, TITLE = {{HyGen}: {G}enerating Random Graphs with Hyperbolic Communities}, AUTHOR = {Metzler, Saskia and Miettinen, Pauli}, LANGUAGE = {eng}, ISSN = {2364-8228}, DOI = {10.1007/s41109-019-0166-8}, PUBLISHER = {Springer}, ADDRESS = {New York, NY}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, JOURNAL = {Applied Network Science}, VOLUME = {4}, EID = {53}, }
Endnote
%0 Journal Article %A Metzler, Saskia %A Miettinen, Pauli %+ Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations %T HyGen: Generating Random Graphs with Hyperbolic Communities : %G eng %U http://hdl.handle.net/21.11116/0000-0005-8E5E-3 %R 10.1007/s41109-019-0166-8 %7 2019 %D 2019 %J Applied Network Science %O ANS Appl Netw Sci %V 4 %Z sequence number: 53 %I Springer %C New York, NY %@ false
[60]
M. Mohanty, M. Ramanath, M. Yahya, and G. Weikum, “Spec-QP: Speculative Query Planning for Joins over Knowledge Graphs,” in Advances in Database Technology (EDBT 2019), Lisbon, Portugal, 2019.
Export
BibTeX
@inproceedings{Mohanty:EDBT2019, TITLE = {{Spec-QP}: {S}peculative Query Planning for Joins over Knowledge Graphs}, AUTHOR = {Mohanty, Madhulika and Ramanath, Maya and Yahya, Mohamed and Weikum, Gerhard}, LANGUAGE = {eng}, ISBN = {978-3-89318-081-3}, DOI = {10.5441/002/edbt.2019.07}, PUBLISHER = {OpenProceedings.org}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {Advances in Database Technology (EDBT 2019)}, EDITOR = {Herschel, Melanie and Galhardas, Helena and Reinwald, Berthold and Fundlaki, Irini and Binning, Carsten and Kaoudi, Zoi}, PAGES = {61--72}, ADDRESS = {Lisbon, Portugal}, }
Endnote
%0 Conference Proceedings %A Mohanty, Madhulika %A Ramanath, Maya %A Yahya, Mohamed %A Weikum, Gerhard %+ External Organizations External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T Spec-QP: Speculative Query Planning for Joins over Knowledge Graphs : %G eng %U http://hdl.handle.net/21.11116/0000-0003-3A7D-1 %R 10.5441/002/edbt.2019.07 %D 2019 %B 22nd International Conference on Extending Database Technology %Z date of event: 2019-03-26 - 2019-03-29 %C Lisbon, Portugal %B Advances in Database Technology %E Herschel, Melanie; Galhardas, Helena; Reinwald, Berthold; Fundlaki, Irini; Binning, Carsten; Kaoudi, Zoi %P 61 - 72 %I OpenProceedings.org %@ 978-3-89318-081-3
[61]
S. Nag Chowdhury, N. Tandon, H. Ferhatosmanoglu, and G. Weikum, “VISIR: Visual and Semantic Image Label Refinement,” 2019. [Online]. Available: http://arxiv.org/abs/1909.00741. (arXiv: 1909.00741)
Abstract
The social media explosion has populated the Internet with a wealth of images. There are two existing paradigms for image retrieval: 1) content-based image retrieval (CBIR), which has traditionally used visual features for similarity search (e.g., SIFT features), and 2) tag-based image retrieval (TBIR), which has relied on user tagging (e.g., Flickr tags). CBIR now gains semantic expressiveness by advances in deep-learning-based detection of visual labels. TBIR benefits from query-and-click logs to automatically infer more informative labels. However, learning-based tagging still yields noisy labels and is restricted to concrete objects, missing out on generalizations and abstractions. Click-based tagging is limited to terms that appear in the textual context of an image or in queries that lead to a click. This paper addresses the above limitations by semantically refining and expanding the labels suggested by learning-based object detection. We consider the semantic coherence between the labels for different objects, leverage lexical and commonsense knowledge, and cast the label assignment into a constrained optimization problem solved by an integer linear program. Experiments show that our method, called VISIR, improves the quality of the state-of-the-art visual labeling tools like LSDA and YOLO.
Export
BibTeX
@online{Nag_arXiv1909.00741, TITLE = {{VISIR}: Visual and Semantic Image Label Refinement}, AUTHOR = {Nag Chowdhury, Sreyasi and Tandon, Niket and Ferhatosmanoglu, Hakan and Weikum, Gerhard}, LANGUAGE = {eng}, URL = {http://arxiv.org/abs/1909.00741}, EPRINT = {1909.00741}, EPRINTTYPE = {arXiv}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, ABSTRACT = {The social media explosion has populated the Internet with a wealth of images. There are two existing paradigms for image retrieval: 1) content-based image retrieval (CBIR), which has traditionally used visual features for similarity search (e.g., SIFT features), and 2) tag-based image retrieval (TBIR), which has relied on user tagging (e.g., Flickr tags). CBIR now gains semantic expressiveness by advances in deep-learning-based detection of visual labels. TBIR benefits from query-and-click logs to automatically infer more informative labels. However, learning-based tagging still yields noisy labels and is restricted to concrete objects, missing out on generalizations and abstractions. Click-based tagging is limited to terms that appear in the textual context of an image or in queries that lead to a click. This paper addresses the above limitations by semantically refining and expanding the labels suggested by learning-based object detection. We consider the semantic coherence between the labels for different objects, leverage lexical and commonsense knowledge, and cast the label assignment into a constrained optimization problem solved by an integer linear program. Experiments show that our method, called VISIR, improves the quality of the state-of-the-art visual labeling tools like LSDA and YOLO.}, }
Endnote
%0 Report %A Nag Chowdhury, Sreyasi %A Tandon, Niket %A Ferhatosmanoglu, Hakan %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society %T VISIR: Visual and Semantic Image Label Refinement : %G eng %U http://hdl.handle.net/21.11116/0000-0005-83CE-F %U http://arxiv.org/abs/1909.00741 %D 2019 %X The social media explosion has populated the Internet with a wealth of images. There are two existing paradigms for image retrieval: 1) content-based image retrieval (CBIR), which has traditionally used visual features for similarity search (e.g., SIFT features), and 2) tag-based image retrieval (TBIR), which has relied on user tagging (e.g., Flickr tags). CBIR now gains semantic expressiveness by advances in deep-learning-based detection of visual labels. TBIR benefits from query-and-click logs to automatically infer more informative labels. However, learning-based tagging still yields noisy labels and is restricted to concrete objects, missing out on generalizations and abstractions. Click-based tagging is limited to terms that appear in the textual context of an image or in queries that lead to a click. This paper addresses the above limitations by semantically refining and expanding the labels suggested by learning-based object detection. We consider the semantic coherence between the labels for different objects, leverage lexical and commonsense knowledge, and cast the label assignment into a constrained optimization problem solved by an integer linear program. Experiments show that our method, called VISIR, improves the quality of the state-of-the-art visual labeling tools like LSDA and YOLO. %K Computer Science, Multimedia, cs.MM,Computer Science, Computer Vision and Pattern Recognition, cs.CV,Computer Science, Information Retrieval, cs.IR
[62]
S. Nag Chowdhury, S. Razniewski, and G. Weikum, “Story-oriented Image Selection and Placement,” 2019. [Online]. Available: http://arxiv.org/abs/1909.00692. (arXiv: 1909.00692)
Abstract
Multimodal contents have become commonplace on the Internet today, manifested as news articles, social media posts, and personal or business blog posts. Among the various kinds of media (images, videos, graphics, icons, audio) used in such multimodal stories, images are the most popular. The selection of images from a collection - either author's personal photo album, or web repositories - and their meticulous placement within a text, builds a succinct multimodal commentary for digital consumption. In this paper we present a system that automates the process of selecting relevant images for a story and placing them at contextual paragraphs within the story for a multimodal narration. We leverage automatic object recognition, user-provided tags, and commonsense knowledge, and use an unsupervised combinatorial optimization to solve the selection and placement problems seamlessly as a single unit.
Export
BibTeX
@online{Nag_arXiv1909.00692, TITLE = {Story-oriented Image Selection and Placement}, AUTHOR = {Nag Chowdhury, Sreyasi and Razniewski, Simon and Weikum, Gerhard}, LANGUAGE = {eng}, URL = {http://arxiv.org/abs/1909.00692}, EPRINT = {1909.00692}, EPRINTTYPE = {arXiv}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, ABSTRACT = {Multimodal contents have become commonplace on the Internet today, manifested as news articles, social media posts, and personal or business blog posts. Among the various kinds of media (images, videos, graphics, icons, audio) used in such multimodal stories, images are the most popular. The selection of images from a collection -- either author's personal photo album, or web repositories -- and their meticulous placement within a text, builds a succinct multimodal commentary for digital consumption. In this paper we present a system that automates the process of selecting relevant images for a story and placing them at contextual paragraphs within the story for a multimodal narration. We leverage automatic object recognition, user-provided tags, and commonsense knowledge, and use an unsupervised combinatorial optimization to solve the selection and placement problems seamlessly as a single unit.}, }
Endnote
%0 Report %A Nag Chowdhury, Sreyasi %A Razniewski, Simon %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T Story-oriented Image Selection and Placement : %G eng %U http://hdl.handle.net/21.11116/0000-0005-83C9-4 %U http://arxiv.org/abs/1909.00692 %D 2019 %X Multimodal contents have become commonplace on the Internet today, manifested as news articles, social media posts, and personal or business blog posts. Among the various kinds of media (images, videos, graphics, icons, audio) used in such multimodal stories, images are the most popular. The selection of images from a collection - either author's personal photo album, or web repositories - and their meticulous placement within a text, builds a succinct multimodal commentary for digital consumption. In this paper we present a system that automates the process of selecting relevant images for a story and placing them at contextual paragraphs within the story for a multimodal narration. We leverage automatic object recognition, user-provided tags, and commonsense knowledge, and use an unsupervised combinatorial optimization to solve the selection and placement problems seamlessly as a single unit. %K Computer Science, Computation and Language, cs.CL
[63]
S. Nag Chowdhury, N. Tandon, and G. Weikum, “Know2Look: Commonsense Knowledge for Visual Search,” 2019. [Online]. Available: http://arxiv.org/abs/1909.00749. (arXiv: 1909.00749)
Abstract
With the rise in popularity of social media, images accompanied by contextual text form a huge section of the web. However, search and retrieval of documents are still largely dependent on solely textual cues. Although visual cues have started to gain focus, the imperfection in object/scene detection do not lead to significantly improved results. We hypothesize that the use of background commonsense knowledge on query terms can significantly aid in retrieval of documents with associated images. To this end we deploy three different modalities - text, visual cues, and commonsense knowledge pertaining to the query - as a recipe for efficient search and retrieval.
Export
BibTeX
@online{Nag_arXiv1909.00749, TITLE = {{Know2Look}: Commonsense Knowledge for Visual Search}, AUTHOR = {Nag Chowdhury, Sreyasi and Tandon, Niket and Weikum, Gerhard}, LANGUAGE = {eng}, URL = {http://arxiv.org/abs/1909.00749}, EPRINT = {1909.00749}, EPRINTTYPE = {arXiv}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, ABSTRACT = {With the rise in popularity of social media, images accompanied by contextual text form a huge section of the web. However, search and retrieval of documents are still largely dependent on solely textual cues. Although visual cues have started to gain focus, the imperfection in object/scene detection do not lead to significantly improved results. We hypothesize that the use of background commonsense knowledge on query terms can significantly aid in retrieval of documents with associated images. To this end we deploy three different modalities -- text, visual cues, and commonsense knowledge pertaining to the query -- as a recipe for efficient search and retrieval.}, }
Endnote
%0 Report %A Nag Chowdhury, Sreyasi %A Tandon, Niket %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T Know2Look: Commonsense Knowledge for Visual Search : %G eng %U http://hdl.handle.net/21.11116/0000-0005-83D2-9 %U http://arxiv.org/abs/1909.00749 %D 2019 %X With the rise in popularity of social media, images accompanied by contextual text form a huge section of the web. However, search and retrieval of documents are still largely dependent on solely textual cues. Although visual cues have started to gain focus, the imperfection in object/scene detection do not lead to significantly improved results. We hypothesize that the use of background commonsense knowledge on query terms can significantly aid in retrieval of documents with associated images. To this end we deploy three different modalities - text, visual cues, and commonsense knowledge pertaining to the query - as a recipe for efficient search and retrieval. %K Computer Science, Information Retrieval, cs.IR
[64]
S. Paramonov, D. Stepanova, and P. Miettinen, “Hybrid ASP-based Approach to Pattern Mining,” Theory and Practice of Logic Programming, vol. 19, no. 4, 2019.
Export
BibTeX
@article{ParamonovTPLP, TITLE = {Hybrid {ASP}-based Approach to Pattern Mining}, AUTHOR = {Paramonov, Sergey and Stepanova, Daria and Miettinen, Pauli}, LANGUAGE = {eng}, ISSN = {1471-0684}, DOI = {10.1017/S1471068418000467}, PUBLISHER = {Cambridge University Press}, ADDRESS = {Cambridge}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, JOURNAL = {Theory and Practice of Logic Programming}, VOLUME = {19}, NUMBER = {4}, PAGES = {505--535}, }
Endnote
%0 Journal Article %A Paramonov, Sergey %A Stepanova, Daria %A Miettinen, Pauli %+ External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T Hybrid ASP-based Approach to Pattern Mining : %G eng %U http://hdl.handle.net/21.11116/0000-0003-0CC4-3 %R 10.1017/S1471068418000467 %7 2019 %D 2019 %J Theory and Practice of Logic Programming %O TPLP %V 19 %N 4 %& 505 %P 505 - 535 %I Cambridge University Press %C Cambridge %@ false
[65]
K. Popat, “Credibility Analysis of Textual Claimswith Explainable Evidence,” Universität des Saarlandes, Saarbrücken, 2019.
Abstract
Despite being a vast resource of valuable information, the Web has been polluted by the spread of false claims. Increasing hoaxes, fake news, and misleading information on the Web have given rise to many fact-checking websites that manually assess these doubtful claims. However, the rapid speed and large scale of misinformation spread have become the bottleneck for manual verification. This calls for credibility assessment tools that can automate this verification process. Prior works in this domain make strong assumptions about the structure of the claims and the communities where they are made. Most importantly, black-box techniques proposed in prior works lack the ability to explain why a certain statement is deemed credible or not. To address these limitations, this dissertation proposes a general framework for automated credibility assessment that does not make any assumption about the structure or origin of the claims. Specifically, we propose a feature-based model, which automatically retrieves relevant articles about the given claim and assesses its credibility by capturing the mutual interaction between the language style of the relevant articles, their stance towards the claim, and the trustworthiness of the underlying web sources. We further enhance our credibility assessment approach and propose a neural-network-based model. Unlike the feature-based model, this model does not rely on feature engineering and external lexicons. Both our models make their assessments interpretable by extracting explainable evidence from judiciously selected web sources. We utilize our models and develop a Web interface, CredEye, which enables users to automatically assess the credibility of a textual claim and dissect into the assessment by browsing through judiciously and automatically selected evidence snippets. In addition, we study the problem of stance classification and propose a neural-network-based model for predicting the stance of diverse user perspectives regarding the controversial claims. Given a controversial claim and a user comment, our stance classification model predicts whether the user comment is supporting or opposing the claim.
Export
BibTeX
@phdthesis{Popatphd2019, TITLE = {Credibility Analysis of Textual Claimswith Explainable Evidence}, AUTHOR = {Popat, Kashyap}, LANGUAGE = {eng}, DOI = {10.22028/D291-30005}, SCHOOL = {Universit{\"a}t des Saarlandes}, ADDRESS = {Saarbr{\"u}cken}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, ABSTRACT = {Despite being a vast resource of valuable information, the Web has been polluted by the spread of false claims. Increasing hoaxes, fake news, and misleading information on the Web have given rise to many fact-checking websites that manually assess these doubtful claims. However, the rapid speed and large scale of misinformation spread have become the bottleneck for manual verification. This calls for credibility assessment tools that can automate this verification process. Prior works in this domain make strong assumptions about the structure of the claims and the communities where they are made. Most importantly, black-box techniques proposed in prior works lack the ability to explain why a certain statement is deemed credible or not. To address these limitations, this dissertation proposes a general framework for automated credibility assessment that does not make any assumption about the structure or origin of the claims. Specifically, we propose a feature-based model, which automatically retrieves relevant articles about the given claim and assesses its credibility by capturing the mutual interaction between the language style of the relevant articles, their stance towards the claim, and the trustworthiness of the underlying web sources. We further enhance our credibility assessment approach and propose a neural-network-based model. Unlike the feature-based model, this model does not rely on feature engineering and external lexicons. Both our models make their assessments interpretable by extracting explainable evidence from judiciously selected web sources. We utilize our models and develop a Web interface, CredEye, which enables users to automatically assess the credibility of a textual claim and dissect into the assessment by browsing through judiciously and automatically selected evidence snippets. In addition, we study the problem of stance classification and propose a neural-network-based model for predicting the stance of diverse user perspectives regarding the controversial claims. Given a controversial claim and a user comment, our stance classification model predicts whether the user comment is supporting or opposing the claim.}, }
Endnote
%0 Thesis %A Popat, Kashyap %Y Weikum, Gerhard %A referee: Naumann, Felix %A referee: Yates, Andrew %+ Databases and Information Systems, MPI for Informatics, Max Planck Society International Max Planck Research School, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society %T Credibility Analysis of Textual Claimswith Explainable Evidence : %G eng %U http://hdl.handle.net/21.11116/0000-0005-654D-4 %R 10.22028/D291-30005 %I Universität des Saarlandes %C Saarbrücken %D 2019 %P 134 p. %V phd %9 phd %X Despite being a vast resource of valuable information, the Web has been polluted by the spread of false claims. Increasing hoaxes, fake news, and misleading information on the Web have given rise to many fact-checking websites that manually assess these doubtful claims. However, the rapid speed and large scale of misinformation spread have become the bottleneck for manual verification. This calls for credibility assessment tools that can automate this verification process. Prior works in this domain make strong assumptions about the structure of the claims and the communities where they are made. Most importantly, black-box techniques proposed in prior works lack the ability to explain why a certain statement is deemed credible or not. To address these limitations, this dissertation proposes a general framework for automated credibility assessment that does not make any assumption about the structure or origin of the claims. Specifically, we propose a feature-based model, which automatically retrieves relevant articles about the given claim and assesses its credibility by capturing the mutual interaction between the language style of the relevant articles, their stance towards the claim, and the trustworthiness of the underlying web sources. We further enhance our credibility assessment approach and propose a neural-network-based model. Unlike the feature-based model, this model does not rely on feature engineering and external lexicons. Both our models make their assessments interpretable by extracting explainable evidence from judiciously selected web sources. We utilize our models and develop a Web interface, CredEye, which enables users to automatically assess the credibility of a textual claim and dissect into the assessment by browsing through judiciously and automatically selected evidence snippets. In addition, we study the problem of stance classification and propose a neural-network-based model for predicting the stance of diverse user perspectives regarding the controversial claims. Given a controversial claim and a user comment, our stance classification model predicts whether the user comment is supporting or opposing the claim. %U https://publikationen.sulb.uni-saarland.de/handle/20.500.11880/28481
[66]
K. Popat, S. Mukherjee, A. Yates, and G. Weikum, “STANCY: Stance Classification Based on Consistency Cues,” 2019. [Online]. Available: http://arxiv.org/abs/1910.06048. (arXiv: 1910.06048)
Abstract
Controversial claims are abundant in online media and discussion forums. A better understanding of such claims requires analyzing them from different perspectives. Stance classification is a necessary step for inferring these perspectives in terms of supporting or opposing the claim. In this work, we present a neural network model for stance classification leveraging BERT representations and augmenting them with a novel consistency constraint. Experiments on the Perspectrum dataset, consisting of claims and users' perspectives from various debate websites, demonstrate the effectiveness of our approach over state-of-the-art baselines.
Export
BibTeX
@online{Popat_arXiv1910.06048, TITLE = {{STANCY}: Stance Classification Based on Consistency Cues}, AUTHOR = {Popat, Kashyap and Mukherjee, Subhabrata and Yates, Andrew and Weikum, Gerhard}, LANGUAGE = {eng}, URL = {http://arxiv.org/abs/1910.06048}, EPRINT = {1910.06048}, EPRINTTYPE = {arXiv}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, ABSTRACT = {Controversial claims are abundant in online media and discussion forums. A better understanding of such claims requires analyzing them from different perspectives. Stance classification is a necessary step for inferring these perspectives in terms of supporting or opposing the claim. In this work, we present a neural network model for stance classification leveraging BERT representations and augmenting them with a novel consistency constraint. Experiments on the Perspectrum dataset, consisting of claims and users' perspectives from various debate websites, demonstrate the effectiveness of our approach over state-of-the-art baselines.}, }
Endnote
%0 Report %A Popat, Kashyap %A Mukherjee, Subhabrata %A Yates, Andrew %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T STANCY: Stance Classification Based on Consistency Cues : %G eng %U http://hdl.handle.net/21.11116/0000-0005-83E2-7 %U http://arxiv.org/abs/1910.06048 %D 2019 %X Controversial claims are abundant in online media and discussion forums. A better understanding of such claims requires analyzing them from different perspectives. Stance classification is a necessary step for inferring these perspectives in terms of supporting or opposing the claim. In this work, we present a neural network model for stance classification leveraging BERT representations and augmenting them with a novel consistency constraint. Experiments on the Perspectrum dataset, consisting of claims and users' perspectives from various debate websites, demonstrate the effectiveness of our approach over state-of-the-art baselines. %K Computer Science, Computation and Language, cs.CL,Computer Science, Artificial Intelligence, cs.AI,Computer Science, Learning, cs.LG
[67]
K. Popat, S. Mukherjee, A. Yates, and G. Weikum, “STANCY: Stance Classification Based on Consistency Cues,” in 2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP 2019), Hong Kong, China, 2019.
Export
BibTeX
@inproceedings{D19-1675, TITLE = {STANCY: {S}tance Classification Based on Consistency Cues}, AUTHOR = {Popat, Kashyap and Mukherjee, Subhabrata and Yates, Andrew and Weikum, Gerhard}, LANGUAGE = {eng}, ISBN = {978-1-950737-90-1}, URL = {https://www.aclweb.org/anthology/D19-1675/}, DOI = {10.18653/v1/D19-1675}, PUBLISHER = {ACM}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, BOOKTITLE = {2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP 2019)}, EDITOR = {Inui, Kentaro and JIng, Jiang and Ng, Vincent and Wan, Xiaojun}, PAGES = {6412--6417}, ADDRESS = {Hong Kong, China}, }
Endnote
%0 Conference Proceedings %A Popat, Kashyap %A Mukherjee, Subhabrata %A Yates, Andrew %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T STANCY: Stance Classification Based on Consistency Cues : %G eng %U http://hdl.handle.net/21.11116/0000-0005-827A-F %U https://www.aclweb.org/anthology/D19-1675/ %R 10.18653/v1/D19-1675 %D 2019 %B Conference on Empirical Methods in Natural Language Processing and 9th International Joint Conference on Natural Language Processing %Z date of event: 2019-11-03 - 2019-11-07 %C Hong Kong, China %B 2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint Conference on Natural Language Processing %E Inui, Kentaro; JIng, Jiang; Ng, Vincent; Wan, Xiaojun %P 6412 - 6417 %I ACM %@ 978-1-950737-90-1
[68]
S. Razniewski, N. Jain, P. Mirza, and G. Weikum, “Coverage of Information Extraction from Sentences and Paragraphs,” in 2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP 2019), Hong Kong, China, 2019.
Export
BibTeX
@inproceedings{D19-1000, TITLE = {Coverage of Information Extraction from Sentences and Paragraphs}, AUTHOR = {Razniewski, Simon and Jain, Nitisha and Mirza, Paramita and Weikum, Gerhard}, LANGUAGE = {eng}, ISBN = {978-1-950737-90-1}, URL = {https://www.aclweb.org/anthology/D19-1000}, PUBLISHER = {ACM}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, BOOKTITLE = {2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP 2019)}, EDITOR = {Inui, Kentaro and JIng, Jiang and Ng, Vincent and Wan, Xiaojun}, PAGES = {5770--5775}, ADDRESS = {Hong Kong, China}, }
Endnote
%0 Conference Proceedings %A Razniewski, Simon %A Jain, Nitisha %A Mirza, Paramita %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T Coverage of Information Extraction from Sentences and Paragraphs : %G eng %U http://hdl.handle.net/21.11116/0000-0005-8265-6 %U https://www.aclweb.org/anthology/D19-1000 %D 2019 %B Conference on Empirical Methods in Natural Language Processing and 9th International Joint Conference on Natural Language Processing %Z date of event: 2019-11-03 - 2019-11-07 %C Hong Kong, China %B 2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint Conference on Natural Language Processing %E Inui, Kentaro; JIng, Jiang; Ng, Vincent; Wan, Xiaojun %P 5770 - 5775 %I ACM %@ 978-1-950737-90-1
[69]
J. Romero, S. Razniewski, K. Pal, J. Z. Pan, A. Sakhadeo, and G. Weikum, “Commonsense Properties from Query Logs and Question Answering Forums,” in CIKM’19, 28th ACM International Conference on Information and Knowledge Management, Beijing China, 2019.
Export
BibTeX
@inproceedings{Romero_CIKM2019, TITLE = {Commonsense Properties from Query Logs and Question Answering Forums}, AUTHOR = {Romero, Julien and Razniewski, Simon and Pal, Koninika and Pan, Jeff Z. and Sakhadeo, Archit and Weikum, Gerhard}, LANGUAGE = {eng}, ISBN = {9781450369763}, DOI = {10.1145/3357384.3357955}, PUBLISHER = {ACM}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, BOOKTITLE = {CIKM'19, 28th ACM International Conference on Information and Knowledge Management}, EDITOR = {Zhu, Wenwu and Tao, Dacheng}, PAGES = {1411--1420}, ADDRESS = {Beijing China}, }
Endnote
%0 Conference Proceedings %A Romero, Julien %A Razniewski, Simon %A Pal, Koninika %A Pan, Jeff Z. %A Sakhadeo, Archit %A Weikum, Gerhard %+ External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society %T Commonsense Properties from Query Logs and Question Answering Forums : %G eng %U http://hdl.handle.net/21.11116/0000-0005-8255-8 %R 10.1145/3357384.3357955 %D 2019 %B 28th ACM International Conference on Information and Knowledge Management %Z date of event: 2019-11-03 - 2019-11-07 %C Beijing China %B CIKM'19 %E Zhu, Wenwu; Tao, Dacheng %P 1411 - 1420 %I ACM %@ 9781450369763
[70]
J. Romero, S. Razniewski, K. Pal, J. Z. Pan, A. Sakhadeo, and G. Weikum, “Commonsense Properties from Query Logs and Question Answering Forums,” 2019. [Online]. Available: http://arxiv.org/abs/1905.10989. (arXiv: 1905.10989)
Abstract
Commonsense knowledge about object properties, human behavior and general concepts is crucial for robust AI applications. However, automatic acquisition of this knowledge is challenging because of sparseness and bias in online sources. This paper presents Quasimodo, a methodology and tool suite for distilling commonsense properties from non-standard web sources. We devise novel ways of tapping into search-engine query logs and QA forums, and combining the resulting candidate assertions with statistical cues from encyclopedias, books and image tags in a corroboration step. Unlike prior work on commonsense knowledge bases, Quasimodo focuses on salient properties that are typically associated with certain objects or concepts. Extensive evaluations, including extrinsic use-case studies, show that Quasimodo provides better coverage than state-of-the-art baselines with comparable quality.
Export
BibTeX
@online{Romero_arXiv1905.10989, TITLE = {Commonsense Properties from Query Logs and Question Answering Forums}, AUTHOR = {Romero, Julien and Razniewski, Simon and Pal, Koninika and Pan, Jeff Z. and Sakhadeo, Archit and Weikum, Gerhard}, LANGUAGE = {eng}, URL = {http://arxiv.org/abs/1905.10989}, EPRINT = {1905.10989}, EPRINTTYPE = {arXiv}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, ABSTRACT = {Commonsense knowledge about object properties, human behavior and general concepts is crucial for robust AI applications. However, automatic acquisition of this knowledge is challenging because of sparseness and bias in online sources. This paper presents Quasimodo, a methodology and tool suite for distilling commonsense properties from non-standard web sources. We devise novel ways of tapping into search-engine query logs and QA forums, and combining the resulting candidate assertions with statistical cues from encyclopedias, books and image tags in a corroboration step. Unlike prior work on commonsense knowledge bases, Quasimodo focuses on salient properties that are typically associated with certain objects or concepts. Extensive evaluations, including extrinsic use-case studies, show that Quasimodo provides better coverage than state-of-the-art baselines with comparable quality.}, }
Endnote
%0 Report %A Romero, Julien %A Razniewski, Simon %A Pal, Koninika %A Pan, Jeff Z. %A Sakhadeo, Archit %A Weikum, Gerhard %+ External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T Commonsense Properties from Query Logs and Question Answering Forums : %G eng %U http://hdl.handle.net/21.11116/0000-0003-FEEE-4 %U http://arxiv.org/abs/1905.10989 %D 2019 %X Commonsense knowledge about object properties, human behavior and general concepts is crucial for robust AI applications. However, automatic acquisition of this knowledge is challenging because of sparseness and bias in online sources. This paper presents Quasimodo, a methodology and tool suite for distilling commonsense properties from non-standard web sources. We devise novel ways of tapping into search-engine query logs and QA forums, and combining the resulting candidate assertions with statistical cues from encyclopedias, books and image tags in a corroboration step. Unlike prior work on commonsense knowledge bases, Quasimodo focuses on salient properties that are typically associated with certain objects or concepts. Extensive evaluations, including extrinsic use-case studies, show that Quasimodo provides better coverage than state-of-the-art baselines with comparable quality. %K Computer Science, Computation and Language, cs.CL,Computer Science, Artificial Intelligence, cs.AI,Computer Science, Databases, cs.DB
[71]
F. M. Suchanek, J. Lajus, A. Boschin, and G. Weikum, “Knowledge Representation and Rule Mining in Entity-Centric Knowledge Bases,” in Reasoning Web -- Explainable Artificial Intelligence, Berlin: Springer, 2019.
Export
BibTeX
@incollection{Suchanek_LNCS11810, TITLE = {Knowledge Representation and Rule Mining in Entity-Centric Knowledge Bases}, AUTHOR = {Suchanek, Fabian M. and Lajus, Jonathan and Boschin, Armand and Weikum, Gerhard}, LANGUAGE = {eng}, ISBN = {978-3-030-31422-4}, DOI = {10.1007/978-3-030-31423-1_4}, PUBLISHER = {Springer}, ADDRESS = {Berlin}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {Reasoning Web -- Explainable Artificial Intelligence}, DEBUG = {author: Krötzsch, Markus; author: Stpanova, Daria}, PAGES = {110--152}, SERIES = {Lecture Notes in Computer Science}, VOLUME = {11810}, }
Endnote
%0 Book Section %A Suchanek, Fabian M. %A Lajus, Jonathan %A Boschin, Armand %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society %T Knowledge Representation and Rule Mining in Entity-Centric Knowledge Bases : %G eng %U http://hdl.handle.net/21.11116/0000-0005-8298-C %R 10.1007/978-3-030-31423-1_4 %D 2019 %B Reasoning Web -- Explainable Artificial Intelligence %E Krötzsch, Markus; Stpanova, Daria %P 110 - 152 %I Springer %C Berlin %@ 978-3-030-31422-4 %S Lecture Notes in Computer Science %N 11810
[72]
H. Su, X. Shen, R. Zhang, F. Sun, P. Hu, C. Niu, and J. Zhou, “Improving Multi-turn Dialogue Modelling with Utterance ReWriter,” in Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics (ACL), Florence, Italy, 2019.
Export
BibTeX
@inproceedings{Su_2019, TITLE = {Improving Multi-turn Dialogue Modelling with Utterance {ReWriter}}, AUTHOR = {Su, Hui and Shen, Xiaoyu and Zhang, Rongzhi and Sun, Fei and Hu, Pengwei and Niu, Cheng and Zhou, Jie}, LANGUAGE = {eng}, URL = {https://www.aclweb.org/anthology/P19-1003}, DOI = {10.18653/v1/P19-1003}, PUBLISHER = {Association for Computational Linguistics}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, BOOKTITLE = {Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics (ACL)}, EDITOR = {Korhonen, Anna and Traum, David and M{\`a}rquez, Llu{\'i}s}, PAGES = {22--31}, ADDRESS = {Florence, Italy}, }
Endnote
%0 Conference Proceedings %A Su, Hui %A Shen, Xiaoyu %A Zhang, Rongzhi %A Sun, Fei %A Hu, Pengwei %A Niu, Cheng %A Zhou, Jie %+ External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations External Organizations External Organizations External Organizations %T Improving Multi-turn Dialogue Modelling with Utterance ReWriter : %G eng %U http://hdl.handle.net/21.11116/0000-0005-6982-2 %U https://www.aclweb.org/anthology/P19-1003 %R 10.18653/v1/P19-1003 %D 2019 %B 57th Annual Meeting of the Association for Computational Linguistics %Z date of event: 2019-07-28 - 2019-08-02 %C Florence, Italy %B Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics %E Korhonen, Anna; Traum, David; Màrquez, Lluís %P 22 - 31 %I Association for Computational Linguistics
[73]
N. Tatti and P. Miettinen, “Boolean Matrix Factorization Meets Consecutive Ones Property,” in Proceedings of the 2019 SIAM International Conference on Data Mining (SDM 2019), Calgary, Canada, 2019.
Export
BibTeX
@inproceedings{Tatti_SDM2019, TITLE = {Boolean Matrix Factorization Meets Consecutive Ones Property}, AUTHOR = {Tatti, Nikolaj and Miettinen, Pauli}, LANGUAGE = {eng}, ISBN = {978-1-61197-567-3}, DOI = {10.1137/1.9781611975673.82}, PUBLISHER = {SIAM}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {Proceedings of the 2019 SIAM International Conference on Data Mining (SDM 2019)}, EDITOR = {Berger-Wolf, Tanya and Chawla, Nitesh}, PAGES = {729--737}, ADDRESS = {Calgary, Canada}, }
Endnote
%0 Conference Proceedings %A Tatti, Nikolaj %A Miettinen, Pauli %+ External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society %T Boolean Matrix Factorization Meets Consecutive Ones Property : %G eng %U http://hdl.handle.net/21.11116/0000-0004-030A-E %R 10.1137/1.9781611975673.82 %D 2019 %B SIAM International Conference on Data Mining %Z date of event: 2019-05-02 - 2019-05-04 %C Calgary, Canada %B Proceedings of the 2019 SIAM International Conference on Data Mining %E Berger-Wolf, Tanya; Chawla, Nitesh %P 729 - 737 %I SIAM %@ 978-1-61197-567-3
[74]
N. Tatti and P. Miettinen, “Boolean Matrix Factorization Meets Consecutive Ones Property,” 2019. [Online]. Available: http://arxiv.org/abs/1901.05797. (arXiv: 1901.05797)
Abstract
Boolean matrix factorization is a natural and a popular technique for summarizing binary matrices. In this paper, we study a problem of Boolean matrix factorization where we additionally require that the factor matrices have consecutive ones property (OBMF). A major application of this optimization problem comes from graph visualization: standard techniques for visualizing graphs are circular or linear layout, where nodes are ordered in circle or on a line. A common problem with visualizing graphs is clutter due to too many edges. The standard approach to deal with this is to bundle edges together and represent them as ribbon. We also show that we can use OBMF for edge bundling combined with circular or linear layout techniques. We demonstrate that not only this problem is NP-hard but we cannot have a polynomial-time algorithm that yields a multiplicative approximation guarantee (unless P = NP). On the positive side, we develop a greedy algorithm where at each step we look for the best 1-rank factorization. Since even obtaining 1-rank factorization is NP-hard, we propose an iterative algorithm where we fix one side and and find the other, reverse the roles, and repeat. We show that this step can be done in linear time using pq-trees. We also extend the problem to cyclic ones property and symmetric factorizations. Our experiments show that our algorithms find high-quality factorizations and scale well.
Export
BibTeX
@online{Tatti_arXiv1901.05797, TITLE = {Boolean Matrix Factorization Meets Consecutive Ones Property}, AUTHOR = {Tatti, Nikolaj and Miettinen, Pauli}, URL = {http://arxiv.org/abs/1901.05797}, EPRINT = {1901.05797}, EPRINTTYPE = {arXiv}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, ABSTRACT = {Boolean matrix factorization is a natural and a popular technique for summarizing binary matrices. In this paper, we study a problem of Boolean matrix factorization where we additionally require that the factor matrices have consecutive ones property (OBMF). A major application of this optimization problem comes from graph visualization: standard techniques for visualizing graphs are circular or linear layout, where nodes are ordered in circle or on a line. A common problem with visualizing graphs is clutter due to too many edges. The standard approach to deal with this is to bundle edges together and represent them as ribbon. We also show that we can use OBMF for edge bundling combined with circular or linear layout techniques. We demonstrate that not only this problem is NP-hard but we cannot have a polynomial-time algorithm that yields a multiplicative approximation guarantee (unless P = NP). On the positive side, we develop a greedy algorithm where at each step we look for the best 1-rank factorization. Since even obtaining 1-rank factorization is NP-hard, we propose an iterative algorithm where we fix one side and and find the other, reverse the roles, and repeat. We show that this step can be done in linear time using pq-trees. We also extend the problem to cyclic ones property and symmetric factorizations. Our experiments show that our algorithms find high-quality factorizations and scale well.}, }
Endnote
%0 Report %A Tatti, Nikolaj %A Miettinen, Pauli %+ External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society %T Boolean Matrix Factorization Meets Consecutive Ones Property : %U http://hdl.handle.net/21.11116/0000-0004-02F0-A %U http://arxiv.org/abs/1901.05797 %D 2019 %X Boolean matrix factorization is a natural and a popular technique for summarizing binary matrices. In this paper, we study a problem of Boolean matrix factorization where we additionally require that the factor matrices have consecutive ones property (OBMF). A major application of this optimization problem comes from graph visualization: standard techniques for visualizing graphs are circular or linear layout, where nodes are ordered in circle or on a line. A common problem with visualizing graphs is clutter due to too many edges. The standard approach to deal with this is to bundle edges together and represent them as ribbon. We also show that we can use OBMF for edge bundling combined with circular or linear layout techniques. We demonstrate that not only this problem is NP-hard but we cannot have a polynomial-time algorithm that yields a multiplicative approximation guarantee (unless P = NP). On the positive side, we develop a greedy algorithm where at each step we look for the best 1-rank factorization. Since even obtaining 1-rank factorization is NP-hard, we propose an iterative algorithm where we fix one side and and find the other, reverse the roles, and repeat. We show that this step can be done in linear time using pq-trees. We also extend the problem to cyclic ones property and symmetric factorizations. Our experiments show that our algorithms find high-quality factorizations and scale well. %K Computer Science, Data Structures and Algorithms, cs.DS,Computer Science, Discrete Mathematics, cs.DM,Computer Science, Learning, cs.LG
[75]
A. Tigunova, A. Yates, P. Mirza, and G. Weikum, “Listening between the Lines: Learning Personal Attributes from Conversations,” 2019. [Online]. Available: http://arxiv.org/abs/1904.10887. (arXiv: 1904.10887)
Abstract
Open-domain dialogue agents must be able to converse about many topics while incorporating knowledge about the user into the conversation. In this work we address the acquisition of such knowledge, for personalization in downstream Web applications, by extracting personal attributes from conversations. This problem is more challenging than the established task of information extraction from scientific publications or Wikipedia articles, because dialogues often give merely implicit cues about the speaker. We propose methods for inferring personal attributes, such as profession, age or family status, from conversations using deep learning. Specifically, we propose several Hidden Attribute Models, which are neural networks leveraging attention mechanisms and embeddings. Our methods are trained on a per-predicate basis to output rankings of object values for a given subject-predicate combination (e.g., ranking the doctor and nurse professions high when speakers talk about patients, emergency rooms, etc). Experiments with various conversational texts including Reddit discussions, movie scripts and a collection of crowdsourced personal dialogues demonstrate the viability of our methods and their superior performance compared to state-of-the-art baselines.
Export
BibTeX
@online{Tigunova_arXiv1904.10887, TITLE = {Listening between the Lines: Learning Personal Attributes from Conversations}, AUTHOR = {Tigunova, Anna and Yates, Andrew and Mirza, Paramita and Weikum, Gerhard}, LANGUAGE = {eng}, URL = {http://arxiv.org/abs/1904.10887}, EPRINT = {1904.10887}, EPRINTTYPE = {arXiv}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, ABSTRACT = {Open-domain dialogue agents must be able to converse about many topics while incorporating knowledge about the user into the conversation. In this work we address the acquisition of such knowledge, for personalization in downstream Web applications, by extracting personal attributes from conversations. This problem is more challenging than the established task of information extraction from scientific publications or Wikipedia articles, because dialogues often give merely implicit cues about the speaker. We propose methods for inferring personal attributes, such as profession, age or family status, from conversations using deep learning. Specifically, we propose several Hidden Attribute Models, which are neural networks leveraging attention mechanisms and embeddings. Our methods are trained on a per-predicate basis to output rankings of object values for a given subject-predicate combination (e.g., ranking the doctor and nurse professions high when speakers talk about patients, emergency rooms, etc). Experiments with various conversational texts including Reddit discussions, movie scripts and a collection of crowdsourced personal dialogues demonstrate the viability of our methods and their superior performance compared to state-of-the-art baselines.}, }
Endnote
%0 Report %A Tigunova, Anna %A Yates, Andrew %A Mirza, Paramita %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T Listening between the Lines: Learning Personal Attributes from Conversations : %G eng %U http://hdl.handle.net/21.11116/0000-0003-FE7F-2 %U http://arxiv.org/abs/1904.10887 %D 2019 %X Open-domain dialogue agents must be able to converse about many topics while incorporating knowledge about the user into the conversation. In this work we address the acquisition of such knowledge, for personalization in downstream Web applications, by extracting personal attributes from conversations. This problem is more challenging than the established task of information extraction from scientific publications or Wikipedia articles, because dialogues often give merely implicit cues about the speaker. We propose methods for inferring personal attributes, such as profession, age or family status, from conversations using deep learning. Specifically, we propose several Hidden Attribute Models, which are neural networks leveraging attention mechanisms and embeddings. Our methods are trained on a per-predicate basis to output rankings of object values for a given subject-predicate combination (e.g., ranking the doctor and nurse professions high when speakers talk about patients, emergency rooms, etc). Experiments with various conversational texts including Reddit discussions, movie scripts and a collection of crowdsourced personal dialogues demonstrate the viability of our methods and their superior performance compared to state-of-the-art baselines. %K Computer Science, Computation and Language, cs.CL
[76]
A. Tigunova, A. Yates, P. Mirza, and G. Weikum, “Listening between the Lines: Learning Personal Attributes from Conversations,” in Proceedings of The World Wide Web Conference (WWW 2019), San Francisco, CA, USA, 2019.
Export
BibTeX
@inproceedings{tigunova2019listening, TITLE = {Listening between the Lines: {L}earning Personal Attributes from Conversations}, AUTHOR = {Tigunova, Anna and Yates, Andrew and Mirza, Paramita and Weikum, Gerhard}, LANGUAGE = {eng}, ISBN = {978-1-4503-6674-8}, DOI = {10.1145/3308558.3313498}, PUBLISHER = {ACM}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {Proceedings of The World Wide Web Conference (WWW 2019)}, EDITOR = {McAuley, Julian}, PAGES = {1818--1828}, ADDRESS = {San Francisco, CA, USA}, }
Endnote
%0 Conference Proceedings %A Tigunova, Anna %A Yates, Andrew %A Mirza, Paramita %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society Databases and Information Systems, MPI for Informatics, Max Planck Society %T Listening between the Lines: Learning Personal Attributes from Conversations : %G eng %U http://hdl.handle.net/21.11116/0000-0003-1460-A %R 10.1145/3308558.3313498 %D 2019 %B The Web Conference %Z date of event: 2019-05-13 - 2019-05-17 %C San Francisco, CA, USA %B Proceedings of The World Wide Web Conference %E McAuley, Julian %P 1818 - 1828 %I ACM %@ 978-1-4503-6674-8
[77]
B. D. Trisedya, G. Weikum, J. Qi, and R. Zhang, “Neural Relation Extraction for Knowledge Base Enrichment,” in Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics (ACL), Florence, Italy, 2019.
Export
BibTeX
@inproceedings{Trisedya_ACL2019, TITLE = {Neural Relation Extraction for Knowledge Base Enrichment}, AUTHOR = {Trisedya, Bayu Distiawan and Weikum, Gerhard and Qi, Jianzhong and Zhang, Rui}, LANGUAGE = {eng}, URL = {https://www.aclweb.org/anthology/P19-1023}, DOI = {10.18653/v1/P19-1023}, PUBLISHER = {Association for Computational Linguistics}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, BOOKTITLE = {Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics (ACL)}, EDITOR = {Korhonen, Anna and Traum, David and M{\`a}rguez, Llu{\'i}s}, PAGES = {229--240}, ADDRESS = {Florence, Italy}, }
Endnote
%0 Conference Proceedings %A Trisedya, Bayu Distiawan %A Weikum, Gerhard %A Qi, Jianzhong %A Zhang, Rui %+ External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations %T Neural Relation Extraction for Knowledge Base Enrichment : %G eng %U http://hdl.handle.net/21.11116/0000-0005-6B08-B %U https://www.aclweb.org/anthology/P19-1023 %R 10.18653/v1/P19-1023 %D 2019 %B 57th Annual Meeting of the Association for Computational Linguistics %Z date of event: 2019-07-28 - 2019-08-02 %C Florence, Italy %B Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics %E Korhonen, Anna; Traum, David; Màrguez, Lluís %P 229 - 240 %I Association for Computational Linguistics
[78]
M. Unterkalmsteiner and A. Yates, “Expert-sourcing Domain-specific Knowledge: The Case of Synonym Validation,” in Joint Proceedings of REFSQ-2019 Workshops, Doctoral Symposium, Live Studies Track, and Poster Track co-located with the 25th International Conference on Requirements Engineering: Foundation for Software Quality (REFSQ 2019) (NLP4RE 2019), Essen, Germany, 2019.
Export
BibTeX
@inproceedings{Unterkalmsteiner_NLP4RE2019, TITLE = {Expert-sourcing Domain-specific Knowledge: {The} Case of Synonym Validation}, AUTHOR = {Unterkalmsteiner, Michael and Yates, Andrew}, LANGUAGE = {eng}, ISSN = {1613-0073}, URL = {urn:nbn:de:0074-2376-8}, PUBLISHER = {CEUR-WS}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {Joint Proceedings of REFSQ-2019 Workshops, Doctoral Symposium, Live Studies Track, and Poster Track co-located with the 25th International Conference on Requirements Engineering: Foundation for Software Quality (REFSQ 2019) (NLP4RE 2019)}, EDITOR = {Dalpiaz, Fabiano and Ferrari, Alessio and Franch, Xavier and Gregory, Sarah and Houdek, Frank and Palomares, Cristina}, EID = {8}, SERIES = {CEUR Workshop Proceedings}, VOLUME = {2376}, ADDRESS = {Essen, Germany}, }
Endnote
%0 Conference Proceedings %A Unterkalmsteiner, Michael %A Yates, Andrew %+ External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society %T Expert-sourcing Domain-specific Knowledge: The Case of Synonym Validation : %G eng %U http://hdl.handle.net/21.11116/0000-0004-02AE-6 %D 2019 %B 2nd Workshop on Natural Language Processing for Requirements Engineering and NLP Tool Showcase %Z date of event: 2019-03-18 - 2019-03-18 %C Essen, Germany %B Joint Proceedings of REFSQ-2019 Workshops, Doctoral Symposium, Live Studies Track, and Poster Track co-located with the 25th International Conference on Requirements Engineering: Foundation for Software Quality (REFSQ 2019) %E Dalpiaz, Fabiano; Ferrari, Alessio; Franch, Xavier; Gregory, Sarah; Houdek, Frank; Palomares, Cristina %Z sequence number: 8 %I CEUR-WS %B CEUR Workshop Proceedings %N 2376 %@ false %U http://ceur-ws.org/Vol-2376/NLP4RE19_paper08.pdf
[79]
M. van Leeuwen, P. Chau, J. Vreeken, D. Shahaf, and C. Faloutsos, “Addendum to the Special Issue on Interactive Data Exploration and Analytics (TKDD, Vol. 12, Iss. 1): Introduction by the Guest Editors,” ACM Transactions on Knowledge Discovery from Data, vol. 13, no. 1, 2019.
Export
BibTeX
@article{vanLeeuwen2019, TITLE = {Addendum to the Special Issue on Interactive Data Exploration and Analytics ({TKDD}, Vol. 12, Iss. 1): Introduction by the Guest Editors}, AUTHOR = {van Leeuwen, Matthijs and Chau, Polo and Vreeken, Jilles and Shahaf, Dafna and Faloutsos, Christos}, LANGUAGE = {eng}, ISSN = {1556-4681}, DOI = {10.1145/3298786}, PUBLISHER = {ACM}, ADDRESS = {New York, NY}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, JOURNAL = {ACM Transactions on Knowledge Discovery from Data}, VOLUME = {13}, NUMBER = {1}, EID = {13}, }
Endnote
%0 Journal Article %A van Leeuwen, Matthijs %A Chau, Polo %A Vreeken, Jilles %A Shahaf, Dafna %A Faloutsos, Christos %+ External Organizations External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations %T Addendum to the Special Issue on Interactive Data Exploration and Analytics (TKDD, Vol. 12, Iss. 1): Introduction by the Guest Editors : %G eng %U http://hdl.handle.net/21.11116/0000-0003-FFD5-E %R 10.1145/3298786 %7 2019 %D 2019 %J ACM Transactions on Knowledge Discovery from Data %V 13 %N 1 %Z sequence number: 13 %I ACM %C New York, NY %@ false
[80]
L. Wang, Y. Wang, G. de Melo, and G. Weikum, “Understanding Archetypes of Fake News via Fine-grained Classification,” Social Network Analysis and Mining, vol. 9, no. 1, 2019.
Export
BibTeX
@article{Wang2019_Understanding, TITLE = {Understanding Archetypes of Fake News via Fine-grained Classification}, AUTHOR = {Wang, Liqiang and Wang, Yafang and de Melo, Gerard and Weikum, Gerhard}, LANGUAGE = {eng}, ISSN = {1869-5450}, DOI = {10.1007/s13278-019-0580-z}, PUBLISHER = {Springer}, ADDRESS = {Cham}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, JOURNAL = {Social Network Analysis and Mining}, VOLUME = {9}, NUMBER = {1}, EID = {37}, }
Endnote
%0 Journal Article %A Wang, Liqiang %A Wang, Yafang %A de Melo, Gerard %A Weikum, Gerhard %+ Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society %T Understanding Archetypes of Fake News via Fine-grained Classification : %G eng %U http://hdl.handle.net/21.11116/0000-0005-789A-7 %R 10.1007/s13278-019-0580-z %7 2019 %D 2019 %J Social Network Analysis and Mining %V 9 %N 1 %Z sequence number: 37 %I Springer %C Cham %@ false
[81]
G. Weikum, J. Hoffart, and F. Suchanek, “Knowledge Harvesting: Achievements and Challenges,” in Computing and Software Science, Berlin: Springer, 2019.
Export
BibTeX
@incollection{Weikum_KnowHarv2019, TITLE = {Knowledge Harvesting: Achievements and Challenges}, AUTHOR = {Weikum, Gerhard and Hoffart, Johannes and Suchanek, Fabian}, LANGUAGE = {eng}, ISBN = {978-3-319-91907-2}, DOI = {10.1007/978-3-319-91908-9_13}, PUBLISHER = {Springer}, ADDRESS = {Berlin}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, BOOKTITLE = {Computing and Software Science}, EDITOR = {Steffen, Bernhard and Woeginger, Gerhard}, PAGES = {217--235}, SERIES = {Lecture Notes in Computer Science}, VOLUME = {10000}, }
Endnote
%0 Book Section %A Weikum, Gerhard %A Hoffart, Johannes %A Suchanek, Fabian %+ Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations External Organizations %T Knowledge Harvesting: Achievements and Challenges : %G eng %U http://hdl.handle.net/21.11116/0000-0005-83B1-E %R 10.1007/978-3-319-91908-9_13 %D 2019 %B Computing and Software Science %E Steffen, Bernhard; Woeginger, Gerhard %P 217 - 235 %I Springer %C Berlin %@ 978-3-319-91907-2 %S Lecture Notes in Computer Science %N 10000
[82]
A. Wisesa, F. Darari, A. Krisnadhi, W. Nutt, and S. Razniewski, “Wikidata Completeness Profiling Using ProWD,” in K-CAP’19, 10th International Conference on Knowledge Capture, Marina del Rey, CA, USA, 2019.
Export
BibTeX
@inproceedings{Wisesa_K-CAP2019, TITLE = {Wikidata Completeness Profiling Using {ProWD}}, AUTHOR = {Wisesa, Avicenna and Darari, Fariz and Krisnadhi, Adila and Nutt, Werner and Razniewski, Simon}, LANGUAGE = {eng}, ISBN = {978-1-4503-7008-0}, DOI = {10.1145/3360901.3364425}, PUBLISHER = {ACM}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {K-CAP'19, 10th International Conference on Knowledge Capture}, EDITOR = {Kejriwal, Maynak and Szekely, Pedro}, PAGES = {123--130}, ADDRESS = {Marina del Rey, CA, USA}, }
Endnote
%0 Conference Proceedings %A Wisesa, Avicenna %A Darari, Fariz %A Krisnadhi, Adila %A Nutt, Werner %A Razniewski, Simon %+ External Organizations External Organizations External Organizations External Organizations Databases and Information Systems, MPI for Informatics, Max Planck Society %T Wikidata Completeness Profiling Using ProWD : %G eng %U http://hdl.handle.net/21.11116/0000-0005-849B-7 %R 10.1145/3360901.3364425 %D 2019 %B 10th International Conference on Knowledge Capture %Z date of event: 2019-11-19 - 2019-11-21 %C Marina del Rey, CA, USA %B K-CAP'19 %E Kejriwal, Maynak; Szekely, Pedro %P 123 - 130 %I ACM %@ 978-1-4503-7008-0
[83]
A. Yates and M. Unterkalmsteiner, “Replicating Relevance-Ranked Synonym Discovery in a New Language and Domain,” in Advances in Information Retrieval (ECIR 2019), Cologne, Germany, 2019.
Export
BibTeX
@inproceedings{Yates_ECIR2019, TITLE = {Replicating Relevance-Ranked Synonym Discovery in a New Language and Domain}, AUTHOR = {Yates, Andrew and Unterkalmsteiner, Michael}, LANGUAGE = {eng}, ISBN = {978-3-030-15711-1}, DOI = {10.1007/978-3-030-15712-8_28}, PUBLISHER = {Springer}, YEAR = {2019}, MARGINALMARK = {$\bullet$}, DATE = {2019}, BOOKTITLE = {Advances in Information Retrieval (ECIR 2019)}, EDITOR = {Azzopardi, Leif and Stein, Benno and Fuhr, Norbert and Mayr, Philipp and Hauff, Claudia and Hiemstra, Djoerd}, PAGES = {429--442}, SERIES = {Lecture Notes in Computer Science}, VOLUME = {11437}, ADDRESS = {Cologne, Germany}, }
Endnote
%0 Conference Proceedings %A Yates, Andrew %A Unterkalmsteiner, Michael %+ Databases and Information Systems, MPI for Informatics, Max Planck Society External Organizations %T Replicating Relevance-Ranked Synonym Discovery in a New Language and Domain : %G eng %U http://hdl.handle.net/21.11116/0000-0004-029B-B %R 10.1007/978-3-030-15712-8_28 %D 2019 %B 41st European Conference on IR Research %Z date of event: 2019-04-14 - 2019-04-18 %C Cologne, Germany %B Advances in Information Retrieval %E Azzopardi, Leif; Stein, Benno; Fuhr, Norbert; Mayr, Philipp; Hauff, Claudia; Hiemstra, Djoerd %P 429 - 442 %I Springer %@ 978-3-030-15711-1 %B Lecture Notes in Computer Science %N 11437