| 摘要 |
PROBLEM TO BE SOLVED: To provide a method, device, and program for automatically classifying texts making it possible to improve classifying accuracy, and to provide a recording medium. SOLUTION: The method for automatically classifying texts uses texts composed of character-string information as the subject of processing, and includes letting a user input a plurality (N) of independent texts, and classifying the texts into a group whose number is smaller than N, based on the similarities among the texts. A process for adapting a cluster analysis processing to the input N texts is carried out, and if two clusters that resemble most are detected as a set of clusters during the process, for a plurality of elements constituting each cluster of the set of clusters, the similarities among the elements are determined among the clusters and compared with a predetermined threshold; when the comparison result does not meet predetermined requirements, the set of clusters is excluded from those whose similarities are to be determined. COPYRIGHT: (C)2004,JPO&NCIPI
|
