Passage Extraction Using Subsequence-Based Query-Sensitive Maximum Cut

Passage extraction is an important component of passage retrieval. Sentence coherence and relevance is two factors mainly considered in the passage extraction. This paper proposes subsequence-based query-sensitive maximum cut algorithm for passage extraction. It incorporates the sentence coherence cut measure and sentence relevance cut measure into normalized cut criterion. And it uses suffix tree model for the text representation and subsequence-based sentence coherence and relevance measure. The experiment results show that our method outperforms some of the existing methods.