中國大陸網民如何看待維穩：2010年至2018年微博的文字探勘研究

Abstract

本文旨在分析中國大陸網民對於維穩文本的探勘研究。為了分析網民對維穩認知的內涵，研究文本取自中國大陸最大規模社群網站「微博」，搜尋並擷取所有「維穩」文章。時間長度共2010到2018九個年度，透過貼文時間趨勢分析、主題分析、word2vec分析、情緒分析等研究方法解析。 經探勘，本文發現數個現象。網民不僅受到網路政策影響減少發文，也會增加部分網民討論網路現況類型的主題。網民在維穩相關文章討論許多議題與事件，反應的情事當中，以強烈譴責情緒評論官僚制度、舉報地方貪官、維穩費議題、基本民生需求議題（醫療、汙染、疫情、強拆）、自由限縮議題（網路現況與實名制）、批評與諷刺中央議題（批評與諷刺當局言論）等。同時也有部分網友高度肯定維穩工作與政績宣傳等。 本文也關注網民對官員與特定事件的反應為何。得知比起中央高官嚴重違紀，網友對地方貪官汙吏更加關注，同時也對不同層級官員的情感反應不一致。網友對禁忌議題如愛國事件與反分裂主義情感也相當不一致，尤其是大陸網民對台獨與領土爭議（釣魚台與南海）有豐富情感表現，中美貿易戰偏向無情感。

The aim of this paper is to examine the opinions of Chinese netizens on the issue of stability maintenance. We fetch “weiwen”(维稳) as keyword to search all posts on Weibo from 2010 to 2018. Through text mining, we used the technique of time trend analysis, topic model analysis, word2vec analysis and sentiment analysis to implement research. We have several findings in this study. First, netizens are influenced by government’s internet censorship policies easily and large amount of posts would be reduced. In the meantime, netizens increase their criticism to the censorship policies as well. Secondly, inside of context, we find out some phenomenon, such as condemnation of the system of bureaucracy, Real-name reporting corrupt officials, the annual state expenditure on stability maintenance, people basic livelihood needs, limitation of freedom, authority’s satire and criticism. In contrast, stability maintenance work and government achievements are appraised and recognized by some netizens. Thirdly, we observe netizen have different sentimental preference on famous politician and major events. Netizens give different response on officials violated discipline or not. And netizens give strong sentiment on patriotism event such as the Diaoyutai dispute and the South China Seas dispute. But it is less sentiment reaction for netizen in Trade War. We keep analysis of high tension of the independence of regions (e.g., Taiwan, Hong Kong, Xinjiang and Tibet). Even less article in these Anti-Secession issues, Taiwanese independence give more strong sentiment reaction than rest of regions.