观点挖掘综述  

Survey of opinion mining

作　　者：陈旻[1] 朱凡微[2] 吴明晖[2] 应晶[1,2] 

机构地区：[1]浙江大学计算机科学与技术学院,浙江杭州310027 [2]浙江大学城市学院计算机科学与工程学系,浙江杭州310015

出　　处：《浙江大学学报（工学版）》2014年第8期1461-1472,共12页Journal of Zhejiang University：Engineering Science

基　　金：浙江省重点科技创新团队项目(2010R50009)

摘　　要：随着互联网技术的飞速发展,博客、社交网络、微博等平台的出现,使得人们在网络上发表个人观点变得更为方便快捷.从海量的数据中提取出消费者、商家、政府等群体需要的信息,并对其进行分析、总结,从而形成有针对性的、可读性强的分析结果.以观点挖掘的流程为线索,从观点提取、极性分析、观点总结等方面对观点挖掘做了较为全面的分析.其中观点提取部分介绍了域依赖性的概念,并例举观点句提取所采用的主要技术;极性分析部分对有监督学习与无监督学习、定性判断极性与定量判断极性进行对比;而观点总结部分则简要介绍了基于方面、观点对比和关键语句等总结的形式.还介绍了在不同的挖掘粒度上所运用的技术,涉及信息检索、自然语言处理、机器学习等.最后,总结了观点挖掘技术中依然存在的具有挑战性的问题,并预测未来观点挖掘领域的研究趋势.With the rapid development of Internet technology, the use of the network in people＇s daily life has become increasingly popular. The appearance of blog, social networking, twitter and some other platforms, makes people express their personal views on the Internet more conveniently. Analyzing these views can provide the most authentic and useful information for consumers, businesses and government. As a result, it becomes particularly important to extract, analyze and summarize information from the massive data, which will be formed as a targeted, readable analysis result for users. In this paper, with the process of opinion mining for clues, we firstly gave a comprehensive analysis on opinion mining from the view of opinion extraction, polarity analyze and opinion summarization, describing the tasks required to complete in each step. Opinion extraction section introduced the concept of domain-dependency, and shows some methods used in extraction. In the polarity analysis section, supervised learning and unsupervised learning, qualitative judgment and quantitative judgment were compared. Summary section gave a brief introduction of several summary forms, such as aspect based, opinion comparison and key works.. Then we introduced different methods on different mining levels, such as document level, sentence level, word level and aspect level, involving techniques like information retrieval, natural language processing, machine learning, etc. Later, we summarized some remaining challenging problems on opinion mining, and proposed.some sentences also contain opinions which unable to determine the polarity future research trends of opinion improved methods respectively. For often be overlooked by researchers. example, sometimes objective At the same time, we still of some complex sentence. In the end of the paper, we predicted the

关 键 词：观点挖掘 观点提取 极性分析 观点总结 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]