Information Gain Ratio meets Maximal Marginal Relevance - A method of Summarization for Multiple Documents

In this paper, we propose a method to make a summary from multiple documents with taking account of comprehensibility and readability. As for comprehensibility, we show an integration of MMR into the termweighting method based on IGR. As for readability, we propose a method to generate a summary based on clustering important sentences according to subtopics and making a keyword list as a very brief summary for each cluster. By the evaluation in NTCIR3 TSC2, we show that the proposed method works well to generate comprehensive summaries when the length of summary is short and the target is a small (7 or less) number of documents.