[Research purpose]The existing early warning methods for the cyber-violence rely heavily on expert knowledge and prior e-vent information,so we propose an early warning method for the cyber-violence event based on Semantic Inconsistency(SI)to realize ef-fective early warning in realistic situation.[Research method]We design a calculation method for SI for the early warning of the cyber-violence event,and integrate a semantic inconsistency monitoring scheme based on anomaly detection in different time sequences,and conduct empirical analysis.We collect"Sina Weibo"data from July to September 2022 for simulated early warning to verify the early-warning capability of the model.[Research conclusion]The average accuracy rate of the early warning reaches 76.57%and the early warning coverage reaches60.92%.The experiment realizes the early warning of up to36 hours in real-world application scenarios.The experiment reveals that a cyber-violence event can have a significantly measurable impact on the topic state of the overall content in the de-velopment stage,and we can realize or enhance the early warning perception ability of the public opinion based on this feature.
关键词
网络暴力/语义不一致性/风险预警/时序异常检测/舆情预警方法
Key words
cyber-violence/semantic inconsistency/risk warning/time series anomaly detection/early warning method
维普
万方数据