A Combination-based Semantic Similarity Measure using Multiple Information Sources

The semantic similarity techniques are interested in determining how much two concepts, or terms, are similar according to a given ontology. This paper proposes a method for measuring semantic similarity/distance between terms. The measure combines strengths and complements weaknesses of existing measures that use ontology as primary source. The proposed measure uses a new feature of common specificity (CSpec) besides the path length feature. The CSpec feature is derived from (1) information content of concepts, and (2) information content of the ontology given a corpus. We evaluated the proposed measure with benchmark test set of term pairs scored for similarity by human experts. The experimental results demonstrated that our similarity measure is effective and outperforms the existing measures. The proposed semantic similarity measure gives the best correlation (0.874) with human scores in the benchmark test set compared to the existing measures