Abstract: Many of the high performing factual question answering systems in the recent TREC (Text Retrieval Conference)`s use a fairly extensive list of surface text patterns. In this study, an automatic surface pattern learning using reformulation rules is proposed. In essence, this is an adaptation of surface pattern learning first proposed by Deepak Ravichandran and Hovy. In our proposed system, predefined sets of representative question and answer patterns, instead of question answer pairs, are used for answer extraction. The performance of the modified system is measured by using two conventional and standard metrics-MRR (Mean Reciprocal Rank) and precision. The system`s performance is also contrasted with that of Hovy`s, so as to elicit improvements due to proposed modifications, using the same metrics.
S.M. Rafizul Haque , Niazur Rahman and Shahabul Islam , 2006. Factual Question Answering by Automatic Surface Pattern Learning Using Reformulation Rules. Asian Journal of Information Technology, 5: 437-441.