基于网络知识百科的情感语义抽取研究

Research on Extraction of Affective Lexical Semantic Using Encyclopedia of Network Knowledge

针对情感词汇语义关系抽取缺乏问题,提出一种简单有效地利用网络知识百科抽取情感词汇语义关系的方法。情感语义关系抽取采用的是递归算法,选用网络百度百科数据源为百度汉语,抽取内容包括情感词汇、情感词汇的同义词和反义词两种情感语义关系。其次,利用抽取出的情感词汇语义关系和现有倾向词典自动扩展标注情感词汇的情感倾向。该方法有效地构建了中文情感词汇语义关系,抽取结果和现有情感词典相比提高了情感词汇数量。同时,基于现有情感词典和抽取的情感词汇间语义关系,实现了快速地扩展情感词语的倾向标注。实验结果表明,抽取获得了2万多个中文情感词汇及其语义关系,并通过情感词汇语义关系实现对抽取词汇的情感倾向扩展标注,准确率达到78.1%。

Aiming at the lack of semantic extraction of affective lexicon,a simple and effective method to extract the semantic relationship of affective lexicon by using knowledge encyclopedia is proposed. Recursion algorithm is adopted to extract the emotional semantic relationship. Baidu encyclopedia online data source is selected as Baidu Chinese,and the extracted content includes two kinds of emotional semantic relationship:emotional vocabulary,synonyms and antonyms of emotional vocabulary. The extracted affective lexicon is labeled the affective tendency based on the extracted semantic relationships and existing tendency dictionaries. The method effectively constructs the semantic relationships of Chinese affective lexicon,improving the number of affective lexicon compared with the existing affective lexicon. At the same time,based on the semantic relationship between the existing sentiment lexicon and the extracted emotional vocabulary,the tendency to quickly expand the emotional words is realized. The experiment shows that more than 2 thousands Chinese affective semantic words and their semantic relationships are extracted,and the emotive tendency of the extracted words is extended and labeled through the semantic relations of emotive words,with an accuracy rate of 78.1%.