The Calculation of Verb Similarity Based on Ternary Morpheme Collocation

Currently, there are about three issues in calculating the similarity of Chinese vocabulary: one is that the calculation of vocabulary similarity generally only focuses on the semantic similarity of words, and the grammatical similarity of words is not paid enough attention; the second is that there is no relevant research to calculate similarity for a part of speech; third, the collocation relationship between words is not fully used. By targeting these three questions, this article locates the verb research object and uses the collocation information between words to calculate the grammatical and semantic similarity of the verb. Firstly, based on the dictionary of verb synonyms, using dependent data to construct the ternary collocation relationship between the verb and its subject-object core morphemes, and embed the ternary collocation information into the vector representation of the verb, finally using the cosine formula to calculate the grammatical and semantic similarity of the verb. According to the different results, the synonyms of a verb are divided into synonyms with the same grammar and semantics, synonyms with similar grammar and semantics, and synonyms with similar semantics. By tagging these three types of labels on the synonyms of verbs, more grammatical and semantic information is provided for clause-level retelling.