抄録
It is important to develop techniques to extract useful information from a large-scale document database. Targeting MEDLINE abstracts, we propose an intelligent system that visualizes relations among abstracts being retrieved by keywords. We characterize relations among abstracts based on “similarity of contents described in abstracts”and “relevance of research contents through authors”. Our system executes a clustering for a retrieved set of abstracts and visualizes the relations among the abstracts in each generated clusters. Our system avails users to trace abstracts of interest efficiently and acquire useful information, e. g. dominant groups or trend of research.
