Extracting Characteristic Sentences from Related Documents

摘要

More and more information is available recently. To find a chance i.e., an important event for decision-making, we have to be prepared for the chance. Recent progress of automatic summarization may contribute to Chance Discovery in that it helps a user read a lot of documents easily and be prepared for the chance. In this paper, we develop a new method for multi-document summarization which extracts a set of characteristic sentences that maximizes the coverage of an original content and minimizes the redundancy of a summary. On top of the summary result, we provide a word cooccurrence graph and show why the result is obtained.