IITD-IBMIRL System for Question Answering Using Pattern Matching, Semantic Type and Semantic Category Recognition

A Question Answering (QA) system aims to return exact answers to natural language questions. While today information retrieval techniques are quite successful at locating within large collections of documents those that are relevant to a user’s query, QA techniques that extract the exact answer from these retrieved documents still do not obtain very good accuracies. We approached the TREC 2007 Question Answering task as a semantics based question to answer matching problem. Given a question we aimed to extract the relevant semantic entities in it so that we can pin-point the answer. In this paper we show that our technique obtains reasonable accuracy compared to other systems.