An Efficient Passage Ranking Technique for a Qa System

Question answering (QA) systems provide an intuitive way of requesting concise information from a given data source. An important stage of such a system is the passage ranking stage, which ranks the possible answers based on their relevance to the question. There has been a lot of previous work on passage ranking, employing lexical, semantic or syntactic methods, but to our knowledge there has been no method that comprehensively combines all 3 features. In this paper, we present a passage ranking technique that leverages lexical, semantic and syntactic features together to rank the answers efficiently and effectively. This paper highlights the differences and improvements of the proposed technique over existing state-ofthe-art techniques like SSTK and IBM Model. The passage ranking technique has been evaluated with TREC QA dataset and is observed to give a significant 26.5% improvement in MRR over the existing stateof-the-art SSTK technique.