Word Semantic Similarity Calculation Based on Cilin and HowNet

In this paper a word semantic similarity calculation method based on the Cilin and HowNet is proposed. In the Cilin part, Cilin after structural transformation is used as the calculation ontology, and the similarity calculation method based on information content is adopted. In the HowNet part, this paper uses the multi-inheritance-based Hownet concept net as the calculation ontology, and adopts the improved similarity calculation method based on information content. Then, according to the distribution of words and the performance of the two algorithms in different scenes, the dynamic weighting strategy considering both Cilin and HowNet is used to calculate the word semantic similarity comprehensively. It has been verified that this method not only expands the range of computable words, but also improves the accuracy of word similarity calculation. In this paper, the parameters are adjusted on the Ruben Stein&Goodenough (RG65) dataset, and the Pearson correlation coefficient between the human judgments in Miller&Charles (MC30) dataset and the calculation results in this paper is 0.944.