基于微博行为的公众社会心态感知
详细信息 查看官网全文
- 英文论文题名：Public Social Mentality Perception Based on Weibo Behavior
- 论文作者：白朔天 ; 朱廷劭
- 英文论文作者：Shuotian Bai ; Tingshao Zhu ; Hubei University of Economics ; Key Laboratory of Behavioral Science ; Institute of Psychology ; Chinese Academy of Sciences
- 年：2015
- 作者机构：湖北经济学院信息工程学院;中国科学院行为科学重点实验室中国科学院心理研究所;
- 论文关键词：社会媒体 ; 网络行为 ; 社会心态 ; 预测
- 英文论文关键词：social media ; network behavior ; social mentality ; prediction
- 会议召开时间：2015-10-16
- 会议录名称：第十八届全国心理学学术会议摘要集——心理学与社会发展
- 英文会议录名称：Proceedings of the 18th National Academic Congress of Psychology
- 语种：中文
- 分类号：C912.6;B841
- 学会代码：ZGXG
- 会议名称：第十八届全国心理学学术会议
- 会议地点：中国天津
- 主办单位：中国心理学会
- 学会名称：中国心理学会
- 页数：2
- 文件大小：776k
- 原文格式：D
- 会议级别：全国
摘要
当前对心态感知方法以主观自评的问卷调查方法为主。问卷调查的方法历史悠久,可以准确获知被试者的心理要素。然而该方法受社会赞许性、测评周期长、成本高等因素影响,无法及时大规模获取用户的相关心理要素。个人心态主导个人行为,反过来也就是说,可通过观测个人行为,对其个人心态进行预测。近些年,我国互联网信息事业得到了迅猛发展,以社会媒体为主的网络社会已成为现实社会的投影,可以满足人们日常生活各方面的需求。在当前的大数据时代下,社会媒体中的网络行为,作为一个极具代表性的个人行为,越发引起学者的关注。社会媒体中的大数据,使得社会心态的感知在客观性、覆盖范围、计算周期和成本投入等方面,弥补了传统研究方法的局限,使通过网络行为分析对用户心态要素的实时可靠预测成为可能。基于以上,本研究从用户网络行为分析入手,尝试预测主导行为的心理动因,即心态。通过社会媒体的跨平台分析,大规模获取用户社会媒体的网上行为痕迹数据,提取了用户网上行为、文本内容和文本情感三方面的多模态特征,建立了用户社会态度多维度心理特征的预测模型。研究表明,模型的正确率超过80%,且达到了心理学意义的中等相关。本研究的主要内容和贡献包括:1)提出了基于情感词典和修正朴素贝叶斯方法的读者情感分类方法,在中文新闻语料库上的实验表明,该方法可对文本的情感进行准确分类;2)提出了基于微博媒体的社会态度感知方法,通过将多任务学习方法与增量拟合方法的融合,提高了从微博数据到社会态度预测的相关性和精度。研究成果应用于北京地区,针对北京公交地铁调价的群体性事件前后,对公众的社会态度进行计算。结果表明群体性事件对公众社会态度具有影响,本研究的模型可对群体性事件前后的社会心态进行即时计算感知。
The current analysis method of social mentality perception is through the subjective self-reported questionnaire survey. Psychological factors of subjects can be accurately informed from questionnaire method with a long history. However, this method is affected by social desirability, self evaluation factors and the long experimental cycle, which makes it unsuitable for real-time and large-scale investigation. The individual mentality leads personal behavior, that is, through the observation of individual behavior, individual mentality can be predicted. In recent years, the information technology industry has been developing rapidly. Social media based social network has become a projection of social reality, which can meets the needs of everyday life. In the current era of big data, network behavior in social media, as a representative of individual behavior, attracted more attention of scholars. The behavior data in social media makes up for the limitations of traditional methods of the perception of social mentality in the aspects of objectivity, coverage, calculation period and the cost of investment. It is possible to real-timely and reliably predict user's mentality elements through the analysis of network behavior. Based on the above, this paper starts with the analysis of network behavior, manages to predict the psychological motivation of the behavior, namely mentality. Through the cross platform analysis of social media, online behavior data traces in large-scale are downloaded. The multimodal features of users' online behavior, text content and text emotion are extracted to establish a prediction model of users of big five personality traits, mental health status and social attitude of multi-dimensional mentality characteristics. Result shows that the prediction model gets a correct rate of over 80% with a middle level of correlation, which has the premise to promote a wide range of computing. The main research contents and contributions of this thesis include: 1) A reader's emotion classification method based on sentiment dictionary and modified Naive Bayes method is presented. Experiments on the Chinese news corpus show that the method can classify text into different emotion categories correctly. 2) A social mentality perception method is proposed through multi task learning and incremental fitting methods which improve the correlation and precision of the prediction from micro blog data to the social mentality. The method is applied to the analysis of social attitudes of micro blog in Beijing. The social attitudes of the public were calculated according to the mass events of the price adjustment of the metro bus in Beijing. The results showed that the social attitudes of the public were influenced by the group events, and our model of this study could calculate the social mentality before and after the mass events in real time.
The current analysis method of social mentality perception is through the subjective self-reported questionnaire survey. Psychological factors of subjects can be accurately informed from questionnaire method with a long history. However, this method is affected by social desirability, self evaluation factors and the long experimental cycle, which makes it unsuitable for real-time and large-scale investigation. The individual mentality leads personal behavior, that is, through the observation of individual behavior, individual mentality can be predicted. In recent years, the information technology industry has been developing rapidly. Social media based social network has become a projection of social reality, which can meets the needs of everyday life. In the current era of big data, network behavior in social media, as a representative of individual behavior, attracted more attention of scholars. The behavior data in social media makes up for the limitations of traditional methods of the perception of social mentality in the aspects of objectivity, coverage, calculation period and the cost of investment. It is possible to real-timely and reliably predict user's mentality elements through the analysis of network behavior. Based on the above, this paper starts with the analysis of network behavior, manages to predict the psychological motivation of the behavior, namely mentality. Through the cross platform analysis of social media, online behavior data traces in large-scale are downloaded. The multimodal features of users' online behavior, text content and text emotion are extracted to establish a prediction model of users of big five personality traits, mental health status and social attitude of multi-dimensional mentality characteristics. Result shows that the prediction model gets a correct rate of over 80% with a middle level of correlation, which has the premise to promote a wide range of computing. The main research contents and contributions of this thesis include: 1) A reader's emotion classification method based on sentiment dictionary and modified Naive Bayes method is presented. Experiments on the Chinese news corpus show that the method can classify text into different emotion categories correctly. 2) A social mentality perception method is proposed through multi task learning and incremental fitting methods which improve the correlation and precision of the prediction from micro blog data to the social mentality. The method is applied to the analysis of social attitudes of micro blog in Beijing. The social attitudes of the public were calculated according to the mass events of the price adjustment of the metro bus in Beijing. The results showed that the social attitudes of the public were influenced by the group events, and our model of this study could calculate the social mentality before and after the mass events in real time.
引文