多策略中文微博观点句识别及情感倾向性判断.pdfVIP

多策略中文微博观点句识别及情感倾向性判断 † † 郭云龙 , 潘玉斌，李国祥，陆阳，余小明，李莉 西南大学计算机与信息科学学院，重庆 400715 † 通讯作者，E-mail: 286207758@，lily@ 摘要 随着社会媒体和社会网络的飞速发展，与微博相关的研究引起了广泛地关注。本文针对中文微博数据(语 句) ，首先采取工具模板进行标注并利用SVM特征分类方法提取微博观点句，然后采用Stanford Parser工具，并 构建情感词典等方法对得到的微博观点句进行情感倾向性分析。本文参加了中国计算机学会(CCF) 自然语言处 理与中文计算会议（NLPCC 2012 ）情感分析词汇语义关系抽取评测。利用会议评测提供的数据，我们得到 观点句准确率和情感分类准确率分别为：78.3%和82.4%1 。 关键词 微博；观点句；情感分析；SVM 中图分类号 Multiple Strategies Based Opinion Sentence Extraction and Sentiment Analysis in Chinese Micro Blog † † GUO Yunlong , PAN Yubing, LI Guoxiang, LU Yang，YU Xiaoming, Li Li School of Computer and Information Science, Southwest University, Chongqing 400715 †Corresponding Author, E-mail: 286207758@，lily@ Abstract With the development of social media and social network, micro-blog has drawn increasing attention from both academia and industry. In this paper, we elaborate on our multiple strategies approach and evaluate it based on the available dataset from NLPCC 2012 . Firstly, syntactic analysis is performed to obtain most of the opinion sentences. This is followed by SVM based approach as a complement to improve the accuracies of opinion sentences extracted. Sentiment analysis is then carried out on these sentences with Stanford Parser tool and the constructed dictionary. The dictionary, initially constructed by search the corpus we obtained from Google, is continually expanded and modified to accommodate the evolution of micro blogging. The experimental results show that the precision of extracting the opinion sentences is 78.3% and t