一种融合性格线索的微博情感分类  

A Weibo sentiment classification combined with the clues of personality

作　　者：贾莉 江涛[1,2] 马宁[1] 孟家豪 JIA Li;JIANG Tao;MA Ning;MENG Jia-hao(Key Laboratory of China's Ethnic Languages and Information Technology of Ministry of Education,Northwest Minzu University,Lanzhou 730030,China;The Silk Road Information Port-Central Asian Information Institute,Northwest Minzu University,Lanzhou 730030,China)

机构地区：[1]西北民族大学中国民族语言文字信息技术教育部重点实验室,甘肃兰州730030 [2]西北民族大学丝绸之路信息港−中西亚信息研究院,甘肃兰州730030

出　　处：《云南大学学报（自然科学版）》2020年第5期870-876,共7页Journal of Yunnan University(Natural Sciences Edition)

基　　金：国家自然科学基金(61762076);2018年度教育部人文社会科学研究青年基金(18YJCZH065)。

摘　　要：现有的文本情感分析模型很少融入性格线索,但不同性格的用户却具有不尽相同的情感表达方式.结合心理学中Big-Five性格模型,提出一种融合性格线索的微博情感分类模型PBiLSTM.该模型将微博句子文本的情感特征与用户性格线索进行融合,增加新的情感判别维度,并利用BiLSTM能够提取文本全局特征的优势,有效提升了模型情感分类的能力.实验结果表明,融合性格线索的微博情感分析模型PBiLSTM的准确率可以达到93.68%,并在多项性能指标上都取得了很好的结果.At present, text sentiment analysis models rarely incorporate personality cues. However, users with different personalities have different emotional expressions. Based on the Big-Five personality model in psychology, this paper proposes a Weibo sentiment classification model PBiLSTM that combines personality cues.The model integrates the sentiment features of Weibo sentence text with the user’s personality cues, thereby adding a new dimension of sentiment classification. At the same time, it uses BiLSTM to extract the advantages of the global features of the text. This method effectively improves the model’s ability to classify emotions. The experimental results show that the accuracy of PBiLSTM method can reach 93.68%, and has achieved good results on multiple performance indicators.

关 键 词：微博 情感分析 性格 Big-Five性格模型 双向长短时记忆网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]