Unsupervised Keyword Extraction Algorithm Integrating Semantic Features
For keyword extraction algorithm,an unsupervised keyword extraction algorithm integrating semantic features is proposed to deal with the lack of text semantic problem,which is intractable for traditional word graph model.This method,which combines the idea of word embedding technology and word graph model to integrate the text semantic information and word order in-formation into the traditional word graph model algorithm.Firstly,by using Word2vec and Doc2vec models to represent words and text respectively,and the word order information of the text is obtained.Then,the semantic similarity between candidate words and text is calculated through the word vector,and then the TextRank algorithm is improved to redistribute the edge weight and initial value between candidate keywords.In addition,the corresponding restart probability matrix and transition probability matrix are con-structed for iterative calculation of candidate word scores and keyword extraction of word graph model.The experimental results show that effectively fusing the semantic information and word information of the text can improve the accuracy of keyword extrac-tion.
万方数据
维普
