基于用户行为的情感分析技术的研究

发布时间：2018-08-02 10:49

【摘要】：随着web2.0应用、移动网络及移动终端的快速普及,社交网络成为人们日常生活的重要组成部分。社交网络上的兴起为情感分析的研究带来了机遇。社交网络的情感分析,有助于服务提供商进行服务优化、精准营销;有助于使用者提高体验、高效消费;有助于政府监管部门进行舆情监测和引导等。目前,许多学者从对短文本的处理和分析入手,或使用情感词典,或提取语义特征进行情感分析工作。而在人类对自身行为的探索和研究中,可以发现行为和情感之间互相影响、互相体现。所以,本文区别于从文本语意角度着手的方法,从网络用户的行为方面进行研究,探索和分析用户行为与其情感倾向之间的关联和规律,并构建基于用户行为的情感倾向分类器。在社交网络中,用户规模庞大,但其行为规范性良好、易于划分和获取,这为研究网络用户行为提供了便利。本文以新浪微博为例,从用户行为的特征入手,研究用户行为在社交网络情感分析工作中的影响和作用。首先就目前的情感分析和用户行为研究现状进行了研究,并系统介绍了社交网络的理论基础和研究情况。随后,根据基于特征的情感分析问题的一般流程,在数据准备阶段进行了数据的抓取、预处理和情感标注等工作;在特征提取阶段,进行了用户行为特征的提取,并通过统计分析和关联规则挖掘方法,研究用户的行为与其情感倾向间的关联关系。最后,对分类技术和贝叶斯及决策树分类器进行了深入系统的研究,结合实际情况,使用朴素贝叶斯算法和C4.5算法构建基于用户行为特征的情感分类模型,并对其进行了实验验证。本文从网络用户的行为方面进行研究,通过一系列研究和实验,证明了社交网络中用户的行为与其情感倾向之间存在一定的联系和规律,从而为进一步的研究打下了基础。
[Abstract]:With the application of web2.0 and the rapid popularization of mobile network and mobile terminal, social network has become an important part of people's daily life. The rise of social networks brings opportunities for emotional analysis. The emotional analysis of social networks is helpful for service providers to optimize services, accurate marketing; help users to improve their experience, efficient consumption; help government regulators to monitor and guide public opinion, and so on. At present, many scholars begin with the processing and analysis of short texts, or use emotional dictionaries or extract semantic features for emotional analysis. In the exploration and study of human behavior, it can be found that behavior and emotion affect each other and reflect each other. Therefore, this paper is different from the approach from the perspective of text semantics, from the point of view of the behavior of network users to explore and analyze the relationship between user behavior and their emotional tendencies and laws. And construct the classifier of emotion tendency based on user behavior. In the social network, the user scale is huge, but its behavior is normative, easy to divide and obtain, which provides convenience for the study of network user behavior. Taking Sina Weibo as an example, this paper studies the influence and function of user behavior in the affective analysis of social network from the characteristics of user behavior. Firstly, the current research status of emotional analysis and user behavior is studied, and the theoretical basis and research situation of social network are systematically introduced. Then, according to the general process of feature-based affective analysis, data capture, preprocessing and affective tagging are carried out in the data preparation stage, and in the feature extraction stage, the user behavior features are extracted. Through statistical analysis and association rule mining, the relationship between user's behavior and emotional tendency is studied. Finally, the classification technology and Bayesian and decision tree classifier are studied deeply and systematically. Combined with the actual situation, the naive Bayesian algorithm and C4.5 algorithm are used to construct the emotion classification model based on user behavior characteristics. It is verified by experiments. Through a series of studies and experiments, this paper proves that there is a certain relationship and law between the user's behavior and their emotional tendency in social network, thus laying a foundation for further research.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.1

【参考文献】