METHOD FOR AUTOMATICALLY CONSTRUCTING AND SEARCHING THESAURUS

PURPOSE: A method for automatically constructing and searching a thesaurus is provided to construct a thesaurus automatically by performing an expression of a centroid with respect to a cluster hierarchically formed as a single word, thereby expressing a hierarchy relation between words by arraying the hierarchies successively. CONSTITUTION: In a construction of a thesaurus automatically using a master file storing index words according to literatures, an index word-literature array is constructed from the master file. An index word-literature array according to orders is constructed based on an appearance frequency of the literatures in the index word-literature array. A complete connection graph-formed cluster is constructed based on a simultaneous appearance frequency of the same literature. A partial connection graph-formed cluster is constructed by connecting clusters, which are not included in the complete connection graph. A centroid of each cluster is extracted by the follow stages. An index having the least literature appearance frequency out of simultaneous appearance words with respect to the lower cluster is decided as a centroid. An index having the least literature appearance frequency except pre-fixed index words out of index words which are appeared more than two times in each literature in the corresponding cluster with respect to the above cluster is decided as a centroid.