Improvement of semantic similarity algorithm based on WordNet

Traditional methods to modeling semantic similarity only compute the common characteristics and uncommon characteristics, without considering the structural relationship between concepts. In this paper, a new semantic similarity method based on information content is proposed. It evolves from feature model, the function λ in the algorithm which defines the relative importance of the uncommon characteristics, depends on the hierarchy in WordNet. In addition, the information content measure also relies on WordNet. Experiments show that, comparing the conventional two similarity measures Resnik and Lin, the proposed method gets more reasonable results with human judgments.