面向民族地区网络内容安全态势感知的跨语言多媒体内容理解

随着互联网的快速发展，网络已经成为社会公共安全态势演化过程中主要的信息传播途径。不同于一般的网络内容安全态势,民族地区网络敏感内容夹杂着民族、宗教等问题，情况错综复杂，影响面大，如果处置不当将造成深远的负面影响。由于现有网络内容安全监测没有对蒙古文的分析，无法准确统计蒙古文信息，为了获得同时获得中文和蒙古文信息并实现有效的内容安全监测，需要在原有分析技术基础上研究跨语言多媒体内容识别和监测技术。通过本项目的研究：(1)提出高效的跨语言多媒体信息发现机制，从庞杂的网络数据中快速抽取各类敏感内容；(2)构建增量深度学习模型，学习跨语言多媒体信息传播过程中的演变，形成准确反映网络敏感内容及事件演化规律的模型；(3)设计网络内容安全态势感知云深度学习监测机制。研究成果对于推动民族地区网络内容安全态势分析与监测具有重要的参考价值。

The internet has become the main way of information dissemination in the evolution of social public security situation. Different from the general security situation of network contents, the sensitive contents of network in ethnic areas are with ethnic and religious issues. The situation is complicated and influential. If we not properly handled the matter, it will cause profound negative impacts. Due to the lack of Mongolian analysis in the existing content security monitoring of network, Mongolian information cannot be accurately counted. In order to obtain both Chinese and Mongolian information and achieve effective content security monitoring, it is necessary to study the cross-language multimedia content identification and monitoring technology based on the original analysis technology. Through the research of this project:(1) We propose an efficient cross-language multimedia information discovery mechanism to quickly extract all kinds of sensitive content from the complex network data. (2) We build an incremental deep learning model to learn the evolution of cross-language multimedia information transmission. The model accurately reflects the evolution of sensitive network contents. (3) We design the deep learning monitoring mechanism for content security awareness of network. The research is instructive to analyze the content security situation of network in minority areas.