An information Content-Based Approach for Measuring Concept Semantic Similarity in WordNet

Computing information content (IC) of a concept is a core issue for semantic similarity measures of IC-based. So far, little works focused on calculating the IC of multiple inheritance nodes. So in this paper, a new IC computing model is proposed to calculate the IC of node (including single inheritance node and multiple inheritances node) in WordNet. This model calculates the IC of the concept through the parameters hypernyms, hyponyms, relative depth, maximum nodes and siblings. Experimental results of “poison” snippet in WordNet taxonomy shows that this model can effectively deal with the cases of single inheritance and multiple inheritance. Meanwhile, the results indicate this model is sensitive to distinguish the IC value of nodes while one of relative depth, hyponym, hypernym or sibling is different. Based on proposed model, a taxonomical semantic similarity measure is proposed to compute the semantic similarity of multiple inheritance nodes. Finally, this paper compares proposed approach with other similarity measures based IC in the given fragment of WordNet classification tree, and the results show that the proposed similarity approach acquires good performance.