An answer passage retrieval strategy for web-based question answering

A passage retrieval strategy for our web-based Question Answering (QA) system is proposed in this paper. We utilize Google to retrieve web documents for answer passage finding. We propose a new method to rewrite the query for passage retrieval. We calculate the relevancy between the query and the passage by combining the term frequency and semantic relevancy. The method has been found effective in the experiment on factoid questions of TREC 2003.