基于影评领域的情感词典构建研究

Construction of Sentiment Lexicon Based on the Field of Film Review

影视评论作为一种典型的带有鲜明主观色彩的文本,其中包含了丰富的情感信息,可以用于情感词典的构建。以采集于豆瓣平台的公开数据集为基础,不断地缩小情感词汇的候选集合,提出一种基于影评领域的情感词典构建方法。利用用户评分的评价方法来确定目标词的极性,同时计算出目标词的情感极性强度,从而构建出基于影评领域的情感词典,最后再引入用户的点赞数进一步优化。结果表明,该方法构建出的情感词典具有较高的精确率,同时,引入点赞数有利于进一步提升构建效果。

As a typical subjective text,film reviews contain a wealth of sentiment information and can be used to build sentiment lexicons. In general,the construction of sentiment lexicons requires existing manual sentiment lexicons,and then specific algorithms or machine learning methods to further expand the lexicon are needed. The paper proposes a method for building sentiment dictionaries based on film reviews by using the public data set collected from Douban Platform as a basis for narrowing down the candidate set of sentiment words. The method adopts user ratings to determine the polarity of the target words,and calculates the sentiment intensity of the target words so as to build a sentiment lexicon based on the field of film review. It finally introduces the number of user likes for further improvement. The result shows that the sentiment lexicon constructed by this method has a high precision rate,and the introduction of user likes facilitates further improvement.