融合事理知识的群体性事件演化预测

Incorporating Factual Knowledge in Mass Events Evolution Prediction

[研究目的]当前,由社会矛盾和冲突所引发的网络群体性事件对公共安全构成了严重威胁,预测群体性事件的演化趋势对提升社会危机的防范与治理能力至关重要。[研究方法]提出了一种融合事理知识的群体性事件演化预测方法。该方法首先利用大语言模型蕴含的丰富事件知识及模型强大的生成能力来构建初始的事理图谱,通过结合真实新闻报道来验证图谱内容;其次,将事理图谱与图神经网络模型相结合,以获得更充分的事件语义表示;最后,基于上下文事件与候选事件间的语义相似度来预测未来可能发生的事件。[研究结论]研究表明,所提方法在事件演化预测的准确性和可解释性方面均显著优于参与比较的其他方法,验证了事理知识融合对揭示群体性事件演化模式的可行性和有效性。

[Research purpose]With the increase of social contradictions and conflicts,the frequent occurrence of mass events poses a serious threat to public security,and prediction of the evolutionary trend of mass events is crucial to enhance the prevention and management of social crises.[Research method]A method for predicting the evolution of mass events by integrating factual knowledge is proposed.The method firstly uses the rich event knowledge contained in the large language model and the powerful generative ability of the model to construct an initial event evolutionary graph,and verifies the content of the graph by combining with real news reports;secondly,it combines the event evolutionary graph with a graph neural network model to obtain a more adequate semantic representation of the event;and lastly,it predicts the possible events in the future based on the semantic similarity between the contextual events and the candidate events.[Research conclusion]The study shows that the proposed method significantly outperforms other methods involved in the comparison in terms of accuracy and interpretability of event evolution prediction,which verifies the feasibility and effectiveness of matter-of-factual knowledge fusion in revealing the evolutionary patterns of mass events.