Multidocument summarization based on concept space

Capturing relevant information is important in supporting decision making. We propose a new summarization method based on cluster analysis, concept space, and statistical approach to extract the essence from a collection of documents. A prototype system has been developed to condense a set of documents into a list of key issues and expands the key issues to form a summary. Cluster analysis and concept space was used as a bridge to connect convergent and divergent processes. Such approach reduces information loss due to vocabulary switching in the summarization process. In the divergent process, it selects the anchored sentences from the original documents to form a summary based on the concept terms generated previously. A user evaluation has been conducted for its usefulness and other performance indices. The results indicate that such approach is promising.