基于分类的微博新情感词抽取方法和特征分析

发布时间：2019-07-02 10:46

【摘要】：情感或情绪分析在舆情分析、商品评论分析、商品推荐等领域应用广泛,而文本中的情感或情绪分析通常以情感词典为基础.人工情感词典虽然准确但构建代价大、难以及时更新,很难适应微博这类新情感词快速更迭的数据.微博平台为新情感词的发布和传播提供了便捷的途径,是新情感词的重要来源.考虑到已有规模较大的人工情感词典及大量包含新情感词的微博数据,在统计、分析、对比中、英两种语言微博中情感词分布差异的基础上,提出了与特定语言无关的基于分类思想的微博新情感词抽取方法cNSEm.cNSEm根据微博数据集和情感词典自动构建训练数据、训练分类器并判别候选词的情感极性,最后采用投票机制确定候选词的情感极性.通过大量而细致的实验,分析了cNSEm在中、英文两种语言的微博数据上的表现、六类特征的作用和用法以及抽取的新情感词对微博情感分类任务的帮助作用.实验结果表明,cNSEm比经典的基于共现和极性传播的方法要好,特别是当考虑中文微博数据集中的名词类情感词时.对cNSEm抽取的新情感词进行了直接和间接两种方法评测,前者利用人工情感词典作参照,后者考察抽取的新情感词对情感分类的帮助作用,从评测指标上看,cNSEm抽取的新情感词与人工情感词典的质量相当,并且cNSEm能适应有较大差异的中、英两个语种.
[Abstract]:Emotional or emotional analysis is widely used in public opinion analysis, commodity review analysis, commodity recommendation and other fields, while emotional or emotional analysis in texts is usually based on emotional dictionaries. Although the artificial emotion dictionary is accurate, it is expensive to build, it is difficult to update it in time, and it is difficult to adapt to the rapid change of new emotional words such as Weibo. Weibo platform provides a convenient way for the release and dissemination of new emotional words, and is an important source of new emotional words. Considering the existing large-scale artificial emotion dictionaries and a large number of Weibo data containing new emotional words, on the basis of statistics, analysis and comparison of the distribution differences of emotional words in Weibo between English and English, a new emotional word extraction method based on classification idea, cNSEm.cNSEm, which is independent of specific language, is proposed to automatically construct training data, train classifiers and distinguish the emotional polarity of candidate words according to Weibo dataset and emotion dictionary. Finally, the emotional polarity of candidate words is determined by voting mechanism. Through a large number of detailed experiments, this paper analyzes the performance of cNSEm on Weibo data in Chinese and English, the function and usage of six kinds of features, and the help of extracted new emotional words to Weibo emotion classification task. The experimental results show that cNSEm is better than the classical method based on co-occurrence and polar propagation, especially when considering the noun emotional words in the Chinese Weibo dataset. The new emotional words extracted by cNSEm are evaluated directly and indirectly. The former uses artificial emotion dictionary as a reference, while the latter examines the helpful effect of the extracted new emotional words on emotion classification. From the evaluation index, the quality of the new emotional words extracted by cNSEm is the same as that of artificial emotion dictionaries, and cNSEm can adapt to Chinese and English languages where there are great differences.
【作者单位】：江西财经大学信息管理学院蒙特利尔大学计算机科学与运筹学系
【基金】：国家自然科学基金(61762042,61363039,61562032) 江西省落地计划项目(KJLD14035) 江西省自然科学基金(20171BAB202021,20152ACB20003)资助~~
【分类号】：TP391.1

【相似文献】