基于词向量的跨领域中文情感词典构建方法

发布时间：2018-03-27 10:48

本文选题：情感分析　切入点：情感词典　出处：《数据采集与处理》2017年03期

【摘要】：情感分析已经成为当今自然语言处理领域的热点问题。对于文本的自动化、半监督式的情感分析研究具有广泛的理论和实用价值。基于情感词典的情感倾向分析方法是文本情感分析的一种重要手段。然而,中文词汇在不同领域中的情感倾向不尽相同,一词多义现象明显。同时,不同领域中的情感词也具有专业性、领域性的特点。针对这些问题,本文提出一种基于词向量相似度的半监督情感极性判断算法(Sentiment orientation from word vector,SO-WV),并依据该算法设计出一种跨领域的中文情感词典构建方法。实验证明,本文所设计的情感词典构建方法能有效地对情感词情感倾向进行判断。算法不仅在不同领域的情感词典建立上具有良好的可移植性,同时还具有专业性、领域性的特点。
[Abstract]:Emotional analysis has become a hot issue in the field of natural language processing. The research of semi-supervised affective analysis has a wide range of theoretical and practical value. Affective tendency analysis based on affective dictionary is an important means of text affective analysis. However, the affective tendency of Chinese vocabulary is different in different fields. At the same time, emotional words in different fields also have professional, domain-specific characteristics. In this paper, we propose a semi-supervised orientation orientation from word vector based on word vector similarity to judge the polarity of Chinese emotion. Based on this algorithm, we design a cross-domain Chinese affective dictionary construction method. The affective dictionary construction method designed in this paper can effectively judge the affective tendency of affective words. The algorithm not only has good portability in the establishment of emotional lexicon in different fields, but also has the characteristics of specialty and domain.
【作者单位】：中国人民大学信息学院;
【基金】：日立企业合作(中国社交媒体的分析服务系统)资助项目中国人民大学科学研究基金(10XNI029)资助项目北京市自然科学基金(4132067)资助项目国家自然科学基金(71271211)资助项目
【分类号】：TP391.1

【相似文献】