Answering List and Other questions

摘要: The importance of Question Answering is growing with the expansion information and text documents on web. Techniques in have significantly improved during last decade especially after introduction TREC track. Most work this field has been done answering Factoid questions. In thesis, however, we present evaluate two approaches to List Other types questions which are as important but not investigated much Although a new research area, them automatically still remains challenge. median F-score systems that participated at TREC-2007 track very low (0.085) while 74% had 0. propose novel approach This based hypothesis answer instances question co-occur within sentences related topic. We use clustering method group candidate answers more often. To pinpoint right cluster, target keywords spies . Using approach, our system placed fourth among 21 teams QA 0.145. introduced TREC-QA retrieve other interesting facts about answered using notion interest marking terms. type questions, extracts, from Wikipedia articles, list terms topic uses extract score document collection where should be found. Sentences then re-ranked universal interest-markers specific top returned possible answers. TREC-2006 tracks. third both years 0.199 0.281 respectively.