ISSN 0439-755X
CN 11-1911/B

心理学报 ›› 2016, Vol. 48 ›› Issue (3): 290-304.doi: 10.3724/SP.J.1041.2016.00290

• 论文 • 上一篇    下一篇



  1. (1北京师范大学心理学院, 北京 100875) (2北京师范大学应用实验心理北京市重点实验室, 北京 100875)
    (3 Deparment of Computer Science and Engineering, The Ohio State University, Columbus, OH 43210 U.S.A)
  • 收稿日期:2015-04-13 出版日期:2016-03-25 发布日期:2016-03-25
  • 通讯作者: 许燕, E-mail:
  • 基金资助:


The characteristics of moral emotions of Chinese netizens towards an anthropogenic hazard: A sentiment analysis on Weibo

YE Yonghao1,2; XU Yan1,2; ZHU Yijie1; LIANG Jiongqian3; LAN Tian1,2; YU Miao1,2   

  1. (1 School of Psychology, Beijing Normal University, Beijing, 100875, China)
    (2 Beijing Key Laboratory of Applied Psychology, Beijing Normal University, Beijing, 100875, China)
    (3 Deparment of Computer Science and Engineering, The Ohio State University, Columbus, OH 43210 U.S.A)
  • Received:2015-04-13 Online:2016-03-25 Published:2016-03-25
  • Contact: XU Yan, E-mail:


本研究采用大数据研究方法, 对爬取的“动车事故”发生后40天内的94,562条相关微博进行情感分析, 以探讨网民对“人祸”的道德情绪特点, 同时对不同群体情绪表达差异进行探讨。结果发现:(1)网民对于动车事故主要表达的道德情绪有:愤怒、鄙视、厌恶、同情和爱。(2)包含不同道德基础的事件与不同的道德情绪相关联; (3)对于愤怒、厌恶和鄙视, 男性普遍有更高的表达倾向和表达强度, 而女性更倾向于表达爱和同情且强度更高; (4)对于爱和同情, 团体VIP用户组表达的可能性和强度都高于其他用户; 个体VIP用户比非VIP用户更可能表达愤怒、鄙视和厌恶, 而团体VIP用户表达这类情绪的强度最小。研究表明, 虚拟网络中人们道德情绪特点依然符合道德基础理论;不同群体在表达道德情绪时的差异性是对道德基础理论相关研究的补充。总言之, 数据挖掘技术和情感分析方法是进行情绪研究的有效手段。

关键词: 道德情绪, 道德基础, 情感分析, 大数据, 温州动车事故


Weibo provides its users a cyber platform to share opinions and show their emotions towards issues at home and abroad. In the process, massive amounts of data are made, and becomes the raw material for sentiment analysis. Previous studies in related fields of computer science and communication focused mainly on developing better sentiment analysis techniques to analyze basic emotions. To add a new perspective, this paper focused on studying the moral emotions expressed toward the “7.23 Wenzhou Train Collision” by Chinese netizens on Weibo. In particular, we analyzed the frequencies of different moral emotions expressed, and related them to the temporal occurrence of different moral events (e.g., statements made by the authority or victims that have moral implications) in the aftermath of the collision, and how different patterns of moral emotions were expressed by different groups including male and female, VIP and non-VIP users.
First of all, we utilized Weibo API to obtain the Weibo Dataset. Specifically, from July 23rd, 2011 to September 1st, 2011 we used several developer IDs to keep grabbing the public timeline, which is a sample of the real time tweets. Then we used a set of keywords to filter out irrelevant tweets and obtain tweets related to the train accident happened on July 23rd. In total, we got 94,562 valid tweets, among which 21,466 tweets contain users’ information. Secondly, we conducted sentiment classification using K-Nearest Neighbor approach based on the training data labeled by 41 experts. After that, all tweets in the dataset were assigned scores from 0 to 5 for each categories of sentiment and the sentiment evolution chart was drawn. Thirdly, we related the knee points of the chart to the moral events happened during the aftermath of the train collision to identify which emotion was evoked by a certain event. Fourthly, we conducted logistic regression and Robust Maximum Likelihood Estimator (MLR) to analyze the difference of emotional expression among different groups.
Results indicated that (1) people showed strong moral anger, contempt, disgust, compassion and love during the moral events that happened in the aftermath of the train accident. (2) The pattern of associations between the moral events and the moral emotions supported the Moral Foundation Theory. (3) Men, in comparison to women, were more likely to express anger, disgust and contempt with higher intensity while women exceeded men at both the likelihood and intensity in expressing compassion and love. (4) For compassion and love, organizational VIP users showed higher likelihood as well as intensity than individual VIP users and non-VIP users. But when it came to anger, disgust and contempt, individual VIP users showed higher likelihood and emotional intensity than non-VIP users and organizational VIP users.

The results are consistent with the Moral Foundation Theory, which postulates that violation of different moral foundation elicits different moral emotions. Moreover, our work is the first one to point out group differences in the expression of moral emotions on Weibo.

Key words: moral emotions, moral foundations, sentiment analysis, big data, Weibo, Wenzhou train collision