高级检索

    面向网络舆情数据的异常行为识别

    Recognition of Abnormal Behavior Based on Data of Public Opinion on the Web

    • 摘要: 社交网络的日益普及和移动设备快捷的网络接入,使得网络舆情的传播十分迅捷,民众对热点话题的关注度和参与度得到很大的提升.网络舆情具有自由性、交互性、多元性、偏差性、突发性等特点,能够左右民众的情感和判断,能推动和改变事件的发展和走向,容易被反对分子利用,已经成为影响社会稳定的重要因素.因此,及时检测、控制并引导舆情的发展具有十分重要的意义.研究关注网络中传播的蕴含有“破坏”、“危险”、“损失”等涉及公共安全或涉及司法公正的行为.根据课题的需要,定义4种关注的异常行为类型:攻击行为、受伤行为、死亡行为、拘捕行为.从数据挖掘和信息抽取的角度研究识别异常行为的方法,首先通过分类器和触发词从海量的数据中过滤出包含异常行为的句子,然后抽取异常行为句中包含的命名实体,最后利用抽取的实体构建异常行为共现网络,为分析人员提供可视化的网络舆情分析方法.

       

      Abstract: With the increasing popularity of the social network, public awareness and participation to hot topics has been much improved, mobile terminal equipment and fast Internet access make the spread of public opinion quickly. Public opinion on the Web has freedom, interactivity, diversity, deviation and burstiness as characteristics, has become an important factor that affects social stability. Therefore, how to timely detect, control and guide the development of public opinion is of great significance to the social stability. This article focuses on the behaviors that spread on the Web and contain “destruction”, “dangerous” and “loss” involves public security or judicial justice, and the behaviors is defined as abnormal behavior. We define the types of abnormal behavior that this article focuses on are aggression, injury, death, and arrests, four categories. From the point of view of information extraction, our method recognizes the abnormal behavior by identifying sentences that contain the abnormal behavior and constructs co-occurrence network of abnormal behavior, with provide the visualization analysis approach of public opinion on the Web.

       

    /

    返回文章
    返回