WebChild: Commonsense Knowledge from the Web

Overview

WebChild is a large collection of commonsense knowledge, automatically extracted and disambiguated from Web contents. WebChild contains triples that connect nouns with adjectives via fine-grained relations like hasShape, hasTaste, evokesEmotion, etc. The arguments of these assertions, nouns and adjectives, are disambiguated by mapping them onto their proper WordNet senses.

Large-scale experiments demonstrate the high accuracy (more than 80 percent) and coverage (more than four million fine grained disambiguated assertions) of WebChild.

Approach

Our method is based on semi-supervised Label Propagation over graphs of noisy candidate assertions. We automatically derive seeds from WordNet and by pattern matching from Web text collections. The Label Propagation algorithm provides us with domain sets and range sets for 19 different relations, and with confidence-ranked assertions between WordNet senses.

• WebChild 2.0: Fine-Grained Commonsense Knowledge Distillation
  Niket Tandon, Gerard de Melo, Gerhard Weikum
  In: ACL Demo 2017.


• Commonsense in Parts: Mining Part-Whole Relations from the Web and Image Tags
  Niket Tandon, Charles Hariman, Jacopo Urbani, Gerhard Weikum, Anna Rohrbach, Marcus Rohrbach
  Accepted: Proc. AAAI 2016.


• Mining Activity Commonsense Knowledge from Hollywood Narratives
  Niket Tandon, Gerard de Melo, Abir De, Gerhard Weikum
  In: Proc. CIKM 2015.
   
• Knowledge Extraction from Movie Scripts
  Niket Tandon, Gerard de Melo, Abir De, Gerhard Weikum
  In: Proc. WWW 2015.
   
• Smarter Than You Think: Acquiring Comparative Commonsense from the Web
  Niket Tandon, Gerard de Melo, Gerhard Weikum (2014)
  Accepted: AAAI 2014, Canada.
   
• WebChild: Harvesting and Organizing Commonsense Knowledge from the Web
  Niket Tandon, Gerard de Melo, Fabian Suchanek, Gerhard Weikum (2014)
  In: Proc. ACM WSDM 2014, New York City, USA.
  Acceptance rate: 18%.

WebBrain: Joint Neural Learning of Large-Scale Commonsense Knowledge