基于购物领域词典扩建的评论情感研究  

Investigation on Sentiment of Reviews with Shopping Field Dictionary Construction

作　　者：吴潇[1] 王磊[1,2] 

机构地区：[1]南京邮电大学宽带无线通信与传感网技术教育部重点实验室,江苏南京210003 [2]东南大学移动通信国家重点实验室,江苏南京210096

出　　处：《计算机技术与发展》2017年第7期194-199,共6页Computer Technology and Development

基　　金：国家自然科学基金资助项目(61201164;61322104;61201165;61271240);南京邮电大学宽带无线通信与传感网技术教育部重点实验室开放研究基金(NYKL201509);东南大学移动通信国家重点实验室开放研究基金(2016D01)

摘　　要：针对购物评论中如何高效提取有用的情感信息,提出了构建领域情感词典进行评论情感分类方法。对购物评论语料进行分词去重,就各领域评论文本进行词性标注,选择词性为名词、形容词及部分其他词性的词语,通过计算该部分词语的PTF-IDF进行排序,设置阈值筛选后得到购物评论语料的领域情感词,从而构建领域情感词典。将该词典作为情感特征应用于购物评论情感分类实验中,并与基于普通情感词典分类方法的性能进行了分析比较。实验结果表明,利用提出方法进行购物评论情感分类的效果,尤其是在分类准确率方面要明显高于基于普通情感词典的情感分类方法,且所提出的方法可适用于各领域的购物评论,有效降低了情感特征空间的维度,具有普适性和可扩展性等优点。Aiming at effectively extracting sentimental information from the shopping reviews, a sentiment classification method has been proposed for online shopping reviews based on the construction of domain emotional dictionary ,which segments words from the reviews corpus, nouns, adjectives and words of some other lexical category in the reviews after part of speech tagging. Sorting by calculating the PTF-IDF of words and setting threshold, selection of domain emotional words has been obtained to construct the domain emotional dic- tionary which has been applied to the sentiment classification of shopping reviews compared with the performance of the classification method based on the general sentiment dictionary. The experimental results show that the proposed method of shopping review sentiment classification is better than one based on general sentiment dictionary, and it can be applied to various fields for effectively reducing the dimension of the feature space, with universality and extensibility.

关 键 词：购物评论情感研究 情感分类 领域情感词典 情感特征 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]