Topic structure extraction apparatus, topic structure extraction program, and computer-readable storage medium storing topic structure extraction program

摘要

PPROBLEM TO BE SOLVED: To detect a plurality of topics, and to automatically generate inter-topic relations and the summary sentence of each topic. PSOLUTION: A text is divided into word units, and a concept base is retrieved so that vectors corresponding to those respective words can be acquired. Then, a text is divided into the groups of segments being the blocks of the same topic from the series of the word vectors, and each segment is considered as the group of the word vectors included in the segment, and the segment groups are hierarchically clustered from such a standard that the segments which are close to each other are the same cluster. As for the respective acquired clusters, a summary sentence characterizing each cluster is extracted from a text included in the cluster, and inter-cluster relations and the summary sentence of each cluster are outputted. PCOPYRIGHT: (C)2005,JPO&NCIPI