面向俄文情感分析的新闻评论语料库建设与应用

Building and Using News Comments Corpus for Russian Sentiment Analysis

从文本中挖掘情感态度是当前计算语言学研究的重点领域和热点方向。论文在分析梳理情感分类方法和俄语情感表达主要手段的基础上,将人的主要情感划分为4大类、19个子类,并基于网络自动构建约730万词的新闻及评论语料库用于实验。为考察俄语情感表达的基本词汇手段,设计一种基于TEI标准的语料情感标注系统,对语料库中热点新闻及其8031条用户评论进行分类标注,所标注语料成功运用到俄语新闻评论的情感自动分析和领域情感词典的构建中,实验结果证明所用方法的有效性和实用性,获得包含6321项情感表达手段的机用词典,后续应用潜力较大。

Mining emotional attitudes from text is the focus and hotspot of current computational linguistics research.After analyzing sentiment classification methods and main means of Russian sentiment expression,this paper divides main human emotions into 4 major categories and 19 sub-categories,and automatically builds a corpus of news and reviews based on the Internet for experi­ment,which contains over 7.3 million words.In order to examine basic vocabulary means of Russian emotional expression,a sentiment labeling system based on the TEI standard was designed.The news in the corpus and its 8031 user comments were clas­sified and labeled.The labeled corpus successfully applied in sentiment analysis of Russian news comments.The experimental results proved the effectiveness and practicability of our methods.A sentiment dictionary containing 6,321 emotion expressions was obtained,which has great potential for subsequent applications.