面向微博可信度评估的辩论图模型

摘要

Due to the mix and uncertainty of the microblog information data, the credibility assessment has difficulty to distinguish. An argumentation directedgraphical model for microblog credibility assessment is presented. From the point of view of argumentation, the model, the argument deduction process is described intuitively by graphical model. In the credibility assessment process, we create the argument node based on the relevant topic corpus, define the rule credibility by the thematic sentiment and potential logical relationship, set the weights on the edges, and represent the defense and attack strength between arguments for this reason. The conclusion credibility is achieved and the argumentation graphical with recursion is evolved according to the relevant algorithm. At last the distinguish information could get the credibility. Experimental results show that this model obtains an accuracy up to 6% in comprehensive indexmeasure compared to the traditional methods.%微博内容具有信息混杂和不确定性等特点,传统可信度判别方法存在一定局限性.因此,该文提出一种面向微博可信度评估的辩论有向图模型,从辩论的角度出发,以图模型直观、形象化地描述了辩论推演过程.通过话题语料构成争议节点,利用争议间的主题情感和潜在逻辑关系定义规则可信度,并设置图中边的权值来代表争议间的防卫和攻击强度.根据相关算法得出结论的可信度,递归进行辩论图演化,得到需判别信息的可信度.实验结果表明该模型比传统方法综合指标值平均提升6%.