by Shuang Kai, Yang Fangchun, Zou Hua, Su Sen
Abstract:
In the semantic web, evaluating the semantic similarity between concepts in a same ontology is a central technique. Traditionally, the distance based approach and the information content based approach are the two major methods. In this paper, on the basis of analyzing these previous approaches, a new method based on ontology hierarchy structure to calculate semantic similarity between concepts is provided. Current similarity methods in the literature only take inheritance relationship into account ignoring composition relationships between ontologies. But we argue that two kinds of relationships together construct a most important part of ontology graph, all of them should be counted when calculating semantic similarities among ontology concepts. We exploit both two relationships to calculate shortest path length (SPL) for improving the quality of similarity measuring. Experimental evaluation against a benchmark set of human similarity ratings demonstrates that the proposed measure outperforms the existing similarity measures and shows a significant correlation to human intuition.
Reference:
An approach for measuring semantic similarity between concepts in semantic web (Shuang Kai, Yang Fangchun, Zou Hua, Su Sen), In IET Conference Publications, 2007.
Bibtex Entry:
@inproceedings{Kai2007,
abstract = {In the semantic web, evaluating the semantic similarity between concepts in a same ontology is a central technique. Traditionally, the distance based approach and the information content based approach are the two major methods. In this paper, on the basis of analyzing these previous approaches, a new method based on ontology hierarchy structure to calculate semantic similarity between concepts is provided. Current similarity methods in the literature only take inheritance relationship into account ignoring composition relationships between ontologies. But we argue that two kinds of relationships together construct a most important part of ontology graph, all of them should be counted when calculating semantic similarities among ontology concepts. We exploit both two relationships to calculate shortest path length (SPL) for improving the quality of similarity measuring. Experimental evaluation against a benchmark set of human similarity ratings demonstrates that the proposed measure outperforms the existing similarity measures and shows a significant correlation to human intuition.},
author = {Kai, Shuang and Fangchun, Yang and Hua, Zou and Sen, Su},
booktitle = {IET Conference Publications},
keywords = {SML-LIB-BIBLIO,lang:ENG,semantic web,semantics},
mendeley-tags = {SML-LIB-BIBLIO,lang:ENG},
number = {529 CP},
pages = {602--608},
title = {{An approach for measuring semantic similarity between concepts in semantic web}},
url = {http://dx.doi.org/10.1049/cp:20070752},
year = {2007}
}