WORD CLASS CREATION PROGRAM, WORD CLASS CREATION APPARATUS, AND WORD CLASS CREATION METHOD; AND SIMILARITY CALCULATION PROGRAM, SIMILARITY CALCULATION APPARATUS, AND SIMILARITY CALCULATION METHOD

摘要

PPROBLEM TO BE SOLVED: To provide a word class creation program and the like capable of creating a group of the same class words in accordance with a target document with very reliably without preparing a word thesaurus for a set of terms appearing in the target document. PSOLUTION: In the program, the similarity is calculated with respect to a combination of extracted terms; a pair of terms comprising two terms are sorted on the basis of the similarity; a graph is created by showing a combination relation by an edge with each term as a node with respect to the selected pair of terms; candidates of dichotomy pattern of the graph obtained by severing a specified edge are extracted; a graph is divided on the basis of an average edge density calculated from the respective candidates; and terms in a configuration node of each of a plurality of graphs in a division result are extracted as a term group of the same class. PCOPYRIGHT: (C)2007,JPO&INPIT