Automatic generation of concept hierarchies using WordNet

摘要

This paper examines and proposes the automatic generation of concept hierarchies using WordNet. Existing researches have mostly explored the utilization of concept hierarchies, but have not addressed the prohibitive cost occurred in building large hierarchies manually. Several studies have examined the automatic generation of concept hierarchies for the numerical type data from a database. However, very little is known about the automatic generation of concept hierarchies for the nominal type data from a database, which is the subject of this paper. We propose the WordNet library method that first eliminates the ambiguity of the senses of nominal data values, constructs the concept hierarchy by overlapping the hypernym of the remaining senses, and lastly adjusts the resultant concept hierarchy to the preference of users. The proposed method is tested with a faculty employment database of a university. The automatic generation of hierarchies turns out to save efforts of experts or designers who build the concept hierarchies, and makes the hierarchy more objectively built than it is manually done.