Word Semantic Similarity Calculation Based on Word2vec

Abstract

In order to solve the problem of poor universality and the absence of contextual information in word similarity calculation based on dictionary, this paper proposes a semantic similarity computation method based on Word2vec. This method improves HowNet and Tongyici Cilin, and also adds the word vector model as a weighing parameter to calculate the word similarity, after compares the similarity of the words by assigning different weights to the three methods. Through experimental comparison, the Pearson coefficient of the algorithm and the artificial value is 0.892, and the method can cover most words so that it can effectively solve the problem of the similarity of the word calculation in the dictionary.