基于HLDA-IDF模型的网络文本主题挖掘研究

Research on Topic Mining of Web Text Based on HLDA-IDF Model

[目的/意义]为了弥补LDA模型建模过程中未考虑到网络文本中文档关注度和质量度这一因素,并增强结果的语义可解释性和主题表示能力,文章提出了一种热度加权的HLDA-IDF的网络文本主题挖掘模型。[方法/过程]本文首先是给出了较为准确的热度定义,并对LDA模型进行热度加权,构建出了HLDA模型,再依据词汇的主题表示能力存在差异这一实际情况,引入TF-IDF算法并改进,构建出HLDA-IDF模型,最后利用实际论坛数据进行实验验证。[结果/结论]实验结果表明该模型的结果语义可解释性和主题表示能力较强。

[ Purpose/significance ] In order to make up the defect that the LDA model does not consider the factor of document attention and quality in web text, and to enhance the semantic interpretation and topic representing ability, this paper proposes a model for web text topic mining based on heat weighted HLDA-IDF. [ Method/process ] This paper first defines heat specifically, and combines with LDA model to construct the HLDA model. Then, as different terms have different topics representing abilities, the paper introduces and improves the TF-IDF algorithm, then builds the HLDA-IDF model. Finally, experiments verify the model by the actual forum data. [ Result/conclusion ] Results show that the proposed model has the better semantic interpretation and topic representing ability.