摘要
[目的/意义]微博作为目前大众获取信息的重要来源,微博中的热门话题则表现出大众关注的焦点所在,如何在热门话题中发现和控制舆情具有重要意义。[方法/过程]文章首先通过网络爬虫程序在新浪微博获取了热点话题微文数据,对热点话题与时事时滞关系进行了统计分析,运用Ward方法对情感进行聚类分析,获取到了情感曲线的8类趋势,最后对不同信息源和情感曲线进行了相关分析,并结合TF-IDF算法识别出了情感极性临界点。[结果 /结论]研究结果表明:公众在社会性话题发生后会及时参与讨论,具有较快的情感反应,公众理性观念形成的关键时期为事件发生的3~10个小时内。话题的情感变化主要源于3种类型的影响,即公众自我感知的主观情感的影响,大量转发的微文情感的影响,自我感知和转发微文两者综合的影响。在话题中可能会出现多个情感极性临界点,而每一次情感极性临界点的出现总是对应能引起情感变化的重大事件或重要微文。
[Purpose/significance]Micro-blog is an important source for public to access to information,and hot topics in micro-blog represent the attention of public. Discovering and controlling public opinions in a hot topic have great significance. [Method/process] This paper focuses on the micro text data of hot topics on Sina Weibo,which are acquired by the web crawler. The relationship between hot topics and time delay of current affairs is statistically analyzed. On the basis of this,the Ward method is used to make clustering analysis of the emotions. Eight types of emotional curves are obtained. And the correlation analysis of different information sources and emotion curves is carried out,and the critical points of emotion polarities are identified by the TF-IDF algorithm. [Result/conclusion] Research results show that the critical period of the formation of the public rational idea is 3 to 10 hours after the event. The public would participate in the discussion right after the social event happened,and they have quick emotional reaction. The emotional changes of topics are mainly due to three types of impacts,namely,the influence of the subjective feelings of the public self perception,the impact of a large number of forwarded micro text emotions,and comprehensive effects of the above two aspects. There may be multiple critical points of emotion polarities in the topic,and the appearance of each critical point is always a major event or important micro text which can cause the change of emotion.
出处
《情报理论与实践》
CSSCI
北大核心
2018年第7期155-160,共6页
Information Studies:Theory & Application
关键词
微博舆情
热点话题
情感识别
聚类分析
micro-blog public opinion
hot topic
emotion recognition
clustering analysis