An improved method for measuring concept semantic similarity combining multiple metrics

Ontology-based semantic similarity measures the similarity between the concepts, which is widely used in information retrieval and semantic web service fields. Existing studies of semantic similarity matching algorithm are mainly focused on computing the semantic distance between concepts, the notion of information content or the overlap of concept attributes. But most of these algorithms calculate semantic similarity in their own way without taking other factors into account. This paper proposes a novel algorithm which combines three factors mentioned above. To avoid unreasonable artificial weight setting, the principal component analysis is used to weigh each factor's contribution to the semantic similarity. The experimental evaluations using WordNet proves that the algorithm presented in this paper improves the accuracy of semantic similarity and the results are more close to human judgment.