【摘要】 近年来,微博平台作为社交媒体载体之一,已经成为新闻信息传播的重要工具。然而,微博平台自身特性决定了其无法提供避免谣言或是虚假信息传递的有效机制。针对这一问题,建立一套完整的算法框架来判断微博的置信度。首先,从不同视角对微博数据提取特征,并将这些多视角的特征通过典型相关分析法(Canonical Correlation Analysis,以下简称CCA)映射到共同子空间中。接下来,从物理学的重力场理论中获得启发,设计一种新的判别学习算法——数据引力场模型(Data Gravitational Field,以下简称DGF)并从大量信息中判别出错误信息或虚假信息。实验表明,这种信息置信度自动检测方法能够达到较高的准确率和召回率。同时,相比较于其它学习算法,数据引力场模型也有更好的表现。更多还原

【Abstract】 As one form of Social Media, microblog has become more popular for news browsing. But it cannot avoid the spread of misinformation and rumors. So an automatic method for judging the credibility of a given set of micro blogs is built. Features of different views have been selected from microblogs, and different views of features are transformed into a common space by Canonical Correlation Analysis(CCA). Then build a framework to identify the misinformation via Data Gravitational Field(DGF). It’s inspired by the theory of the gravitational field of physics. The automatic detection method is able to obtain higher detection accuracy as demonstrated by experiments. At the same time, through a comparative analysis, Data Gravitational Field is also superior to other learning algorithms.更多还原