基于PhraseLDA模型的主题短语挖掘方法研究

Topical Phrase Mining Based on the PhraseLDA Model

[目的/意义]以主题短语识别为研究对象,提出基于PhraseLDA模型的主题短语挖掘方法,为快速理解文本内容、准确抽取文本主题提供借鉴思路。[方法/过程]对低频词进行量化定义,提出一种合理的短语重要度计算方法,最终利用PhraseLDA主题模型推理出主题短语。[结果/结论]实验结果表明该方法在多种数据集中挖掘出的主题短语质量较高,主题一致性较强。

[ Purpose/significance] Taking topical phrase identifying as the research object, this paper provides a method for topical phrase mining based on the Phrase LDA model to understand text content quickly and extract text topic accurately. [ Method/process] This paper provides a quantitative definition of the high frequency words and puts forward a reasonable formula of the phrase importance. We reason out the topical phrases combining the PhraseLDA model finally. [ Result/conclusion] Experimental result shows that the quality of phrases extracted from a variety of data sets is higher and the topical consistency is stronger.