融合Log-Likelihood与TextRank的关键词抽取研究

Research on Keyword Extraction of Fusion Log-Likelihood and TextRank

为了解决TextRank算法的初始值赋权问题,提高关键词抽取准确率,引入Log-Likelihood算法。通过与参考语料库词频进行对比,为词条的初始权重赋值,将不需要外部语料的TextRank和需要外部语料的Log-Likelihood进行融合、计算。实验结果表明,融合后的TextRank-LL算法优于TextRank算法。

In order to solve the initial value of TextRank algorithm, we can improve the accuracy of keyword extraction. The Log-Likelihood algorithm is introduced to compute the initial weight of the term by comparing with the observed word frequency of the corpus. The TextRank without external corpus and the Log-Likelihood which requires external corpus are merged and calculated. Experimental results show that the fusion TextRank-LL algorithm is superior to the TextRank algorithm.