User-focused Automatic Document Summarization using Non-negative Matrix Factorization and Pseudo Relevance Feedback

This paper proposes an automatic document summarization method using the pseudo relevance feedback (PRF) and the non-negative matrix factorization (NMF) to extract sentences relevant to a user's interesting for user-focused summary. The proposed method can improve the quality of document summaries because the inherent structure of the documents are well reflected by using the semantic features and the semantic variables calculated by NMF. Also it can provide an automatic relevance judgment on query expansion without the intervention of user. The experimental results demonstrate that the proposed method achieves better performance the other methods.