社交网络中敏感舆情事件演化趋势的引导与成效评估方法研究

Online Social Network has become the important place for public opinions propagation, which are usually with clear ideas, widely spreading and bad influence. Research on the methods for guiding their changing trends and reducing their influence are a hot and difficult problem in the area of public opinion monitoring. Public opinion guidance is defined as according to its subjective information to generate the corresponding guidance content automatically. And then employ the appropriate guidance strategy to change the subjective understanding of the target audiences. Focus on the existing difficulties, we employ the depth neural network to achieve the goal of subjective information representation. Following the methods for subjective information recognition in cross modal data, such as text, pictures and videos, are proposed. Secondly, the guidance contents are generated automatically based on the characteristics of public opinion, such as time, environment and other factors. Furthermore, we employ the multi-stage guidance strategy to guide the development trends of public opinion at the right time. Finally, we proposed methods based on information fusion to evaluate the effectiveness of public opinion guidance. The results will provide technical support for monitoring the changing trends of public opinion, and contribute to the construction of a harmonious online society.

社交网络已经成为舆情对抗的重要阵地，其中的舆情具有观点明确、传播范围广、影响恶劣等特点，如何引导舆情发展态势以降低其危害是目前舆情监控研究的重点和难点。舆情引导是指针对舆情中的主观信息，自动生成相应的引导内容，采取合适的引导策略，以改变目标受众的主观认识。针对其中存在的难点问题，本项目首先提出了基于深度神经网络的类人脑主观信息表示方法，并在此基础上实现文本、图片、视频等跨模态数据中主观信息的识别。随后，提出了融合舆情事件要素的内容自动生成方法，所生成的文本符合舆情在主题、时间、环境等因素等特征，易被引导目标接受。再者，提出了多阶段拟人化的引导策略，选择合适的时机介入引导，由先期引导和后期疏导为主要目标实现舆情发展趋势的控制。最后，提出了信息融合框架下的舆情引导成效评估方法，并据此实现引导策略的修正和有效性验证。项目研究成果将对舆情发展态势的引导提供技术支撑，为营造和谐的线上社会做出贡献。