基于概念的词汇情感倾向识别方法

Semantic orientation computing based on concepts

词汇的语义倾向是文本倾向性分析的基础课题.现有的词汇语义倾向计算通常是以词汇为基准,而词是包括了多种不同情感倾向概念的粒度范畴,影响分析的精度和效率.据此,提出在更细的粒度下,利用HowNet工具中的"概念"进行倾向性分析,设计了基于概念的语义倾向计算方法.该方法使用聚类的概念,利用K-MEDOIDS算法寻找基准概念.实验结果表明,基于概念的方法较传统基于词汇的方法准确率更高.

The semantic orientation of words is the foundation of sentiment analysis.Current methods to compute semantic orientation of words are mostly based on reference words,while words belonging to the granularity category,including various sentiment orientation concepts,affect the analytical precision and efficiency.In this paper,a new method of semantic orientation computing was proposed based on the reference concepts using the HowNet tool to analyze the tendency.The clustering algorithm K-Mediods was used to search for the reference concepts.The experimental results show that the concept-based method outperforms the word-based method.