基于Word2Vec和TextRank的时政类新闻关键词抽取方法研究

Research of Keyword Extraction of Political News Based on Word2Vec and TextRank

[目的/意义]旨在为时政类新闻关键词抽取提供参考。[方法/过程]基于融合Word2Vec和TextRank算法,在研究时政类新闻文本特征基础上,利用政治重点词库修订文本词语的初始权重,结合上下文关系确定词语之间的连接关系,并基于Word2Vec模型构建概率转移矩阵,提出改进的Word2Vec和TextRank算法。[结果/结论 ]运用改进的Word2Vec和TextRank算法对时政类新闻关键词进行抽取,其准确率、召回率和F值均优于传统TextRank算法及普通的融合Word2Vec和TextRank算法,抽取效果更好。

[Purpose/significance]The paper is to provide reference for keyword extraction of political news. [Method/process]The paper bases on the fusion algorithm of Word2Vec and TextRank, and the research of text features of political news, uses political key thesaurus to revise the initial weights of the text words, combines the context relation to determine the connection relation between words, constructs the Word2Vec model-based probability transfer matrix, and proposes the improved algorithm of Word2Vec and TextRank. [Result/conclusion]With the improved algorithm of Word2Vec and TextRank, the accuracy, recall rate and F value of the extracted results are better than the traditional TextRank algorithm and the common fusion algorithm of Word2Vec and TextRank, thus its extraction effect is the best.