Mining Textual Answers with Knowledge-Based Indicators

This paper describes a knowledge-based methodologyof miningtextual answersinto large collections of texts. Wepresent SOMBRERO,a knowledge processing module implemented in the LASSOquestion answering system. The module bootstraps question taxonomies,used for the search and validation of answers. Answerevaluation is producedthrough an expandableset of abductionrules. Background Whenpeople use computer-based tools to find answers to general questions, they often are faced with a daunting list of search results, or "hits" returned by a search engine. To take a step closer to information retrieval rather than documentretrieval, this year the Text REtrieval Conference (TREC)has initiated an experimental track: the Question/Answering (Q/A) track, whose aim is to foster research in the area of textual Q/A,especially when pursued in a domain independent manner. The TRECQ/A Track specifies two restrictions for the questions. First, questions should have an exact answer that occurs in ...

• Type: Conference
• Year: 2000
• Where: FLAIRS
• Authors: Sanda M. Harabagiu, Marius Pasca
• Keywords: Artificial Intelligence, FLAIRS 2000, Knowledge Processing Module, LASSOquestion Answering System, Module Bootstraps Question