【摘要】 网络暴力识别与多维度分析对网络舆情态势感知和管控具有十分重要的意义，当前的网络暴力研究主要集中在用户负面情感分析、舆情危机治理及外部网络生态优化等方面，缺乏对网络暴力的定量分析研究，无法在复杂的网络环境中及时感知网络暴力发展态势和组成结构。通过分析网络暴力在文本中的存在形式和结构特征，提出了一种基于文本语义的网络暴力分析方法(text semantic based approach for cyber violence analysis, TSCA)。该方法首先运用互信息理论创建暴力领域情感词典，根据暴力领域情感词典和语义环境从评论语料库中生成负面情感词组集；然后，通过卡方检验对负面情感词组集进行暴力特征筛选，生词暴力词组集；最后，从文本和用户角度对网络暴力进行定量计算和多维度分析。在真实的网络热点事件评论文本数据集上与其他方法进行了对比，实验结果表明：该方法达到了良好的网络暴力特征识别效果。更多还原

【Abstract】 Cyber violence identification and multidimensional analysis are of great significance to the situation awareness and control of network public opinion. The current research on cyber violence mainly focuses on the analysis of users’ negative emotions, the governance of public opinion crisis and the optimization of external network ecology, which is unable to sense the development trend and composition structure of cyber violence in a complex network environment. By analyzing the form and structure of cyber violence in text, a text semantic based approach for cyber violence analysis(TSCA) is proposed. Firstly, a violence domain emotion dictionary is created based on mutual information theory, and a negative emotion phrase set is generated from the comment corpus according to the violence domain emotion dictionary and semantic environment. Then, through χ~2 test, the violence features set is screened based on the negative emotion phrase set. Finally, quantitative calculation and multi-dimensional analysis are carried out from the perspective of text and users. The TSCA approach is compared with other methods on the real text data set of network hot event reviews. The experimental results show that the proposed approach achieves a better performance on cyber violence feature recognition.更多还原