TPDBlearn is a system which supports querying and learning over data that is valid during a given time-interval and with a specific probability only. Over this kind of data the system supports queries formulated as deduction rules. Optionally, the user can supply consistency constraints along with the query which then alter the query answers' probabilities. A further feature is learning of tuple probabilities. Given queries labeled by target probabilities, the system can update or learn new probabilities of the tuples such that the labels' target probabilities are met.
If you have any questions or comments please write an email to Maximilian Dylla.
You can download the current version of the source code as of March 22nd, 2014 here. It is licensed under a Creative Commons Attribution-ShareAlike 3.0 Unported License.
Maximilian Dylla, Iris Miliaraki, Martin Theobald
A Temporal-Probabilistic Database Model for Information Extraction (PDF, Bibtex)
Proceedings of the VLDB Endowment, Volume 6, Issue 14, 2013
To be presented at International Conference on Very Large Database Systems, Hangzhou, China, 2014