With Queripidia we want to answer questions located on the fringe of Wikipedia. Given a large, general-purpose knwoledge base and a large corpus of web documents, the goal is to consolidate information from both sources to jointly extract material to aswer the given query with an automatically compiled query-specific Wikipedia.
This is a first prototype to demonstrate the use of such automatic Wikipedia construction, based on queries from TREC Web track 2013. Textual material stems from the Clueweb12 Category A corpus. The knowledge base is taken from a Wikipedia/Freebase snapshot from 2012.
Choose a query. Relevant entities and types will be listed. Click on the entity for an explanation on WHY it is found to be relevant.
Laura Dietz, Michael Schuhmacher, Simone Paolo Ponzetto. Queripidia: Query-specific Wikipedia Construction. NIPS workshop on Automated Knowledge Base Construction (AKBC) 2014.
Jeffrey Dalton, Laura Dietz, James Allan. Entity query feature expansion using knowledge base links. SIGIR 2014.