LDA与词汇链相结合的主题短语抽取方法  

Method of Extracting Topic Phrase Based on LDA and Lexical Chain

作　　者：张小鹏 吕学强[1] 李卓[1] 徐丽萍[2] ZHANG Xiao-peng;LV Xue-qiang;LI Zhuo;XU Li-ping(Beijing Key Laboratory of Internet Culture and Digital Dissemination Research,Beijing Information Science and Technology University,Beijing 100101,China;Beijing Research Center of Urban System Engineering,Beijing 100089,China)

机构地区：[1]北京信息科技大学网络文化与数字传播北京市重点实验室,北京100101 [2]北京城市系统工程研究中心

出　　处：《小型微型计算机系统》2018年第11期2457-2463,共7页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61671070)资助;北京成像技术高精尖创新中心项目(BAICIT-2016003)资助;国家社会科学基金重大项目(14@ZH036)资助;国家语委重点项目(ZDll35-53)资助;国家语委重大课题项目(ZDA125-26)资助.

摘　　要：主题短语能够有效地概括文献主要内容,针对性地反映文献的核心思想.为了改进技术文献的主题抽取效果,在研究LDA主题模型和词汇链的基础上,提出了LDA模型与词汇链相结合的主题短语抽取方法,减小了利用单一LDA模型抽取文献主题中存在的主题漂移问题;在主题短语抽取中,利用对数似然比解决了词汇链构造过程中知识库未收录词的词语相关度的计算问题.实验结果表明,该方法相对于传统方法的主题抽取效果具有明显的改善,且准确率和召回率均有所提升.对信息检索、文本聚类等工作具有十分重要的支撑作用.The topic phrases can effectively summarize the main contents of the literature and reflect the core ideas of the literature objectively. In order to improve the effect of extracting topic from the technical literature,the method of topic extraction based on LDA model and lexical chain is proposed on the basis of studying the LDA topic model and lexical chain, It has reduced the problem of topic drift in the literature topic extraction by using a single LDA model. In the process of topic phrase extraction, the logarithmic likelihood ratio is used to solve the problem of calculating correlation of words which is not included in the knowledge database. Experimental results show that the proposed method has obviously achieved better result than traditional methods in terms of theme extraction, and the accuracy and recall rate have been improved. It plays a very important supporting role in information retrieval and text clustering.

关 键 词：LDA模型 词汇链 WORDNET 主题短语 对数似然比 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]