Semantical-coordinate Terms Detection from Hierarchical Knowledge Using Web Snippets

Abstract In this paper, we describe a method to detect semantical-coordinate terms. We often use semantical-coordinate terms as objects of comparative validation and examples for the given term, and then linguistic expressions and knowledge processing are enriched. Semantical-coordinate terms should be hyponyms of a same hypernym, and their usage and concepts should be similar. However hierarchical knowledge is useful to view concepts, hierarchically coordinate terms are sometimes inappropriate as semantical-coordinate terms for our suppositions, because some of them might have multiple hypernyms and general perceptions of the terms are not taken in the consideration. On the other hand, using only Web context to detect semantical-coordinate terms, concepts of the terms are not taken in the consideration while public perceptions and their usages might be incorporated. Therefore we propose hybrid method using both hierarchical knowledge and Web snippets. We conducted bench scale tests to detect semantical-coordinate terms of some terms and discuss about the results in this paper. Through the tests and discussions, we confirmed that the semantical-coordinate terms detected by our proposed method were not only hierarchically but also semantically and intuitively appropriate.