面向微博可信度评估的辩论图模型

发布时间：2018-05-14 22:37

  本文选题：辩论 + 争议节点　； 参考：《电子科技大学学报》2017年02期

【摘要】：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%。
[Abstract]:The content of Weibo is characterized by mixed information and uncertainty, and there are some limitations in the traditional method of credibility judgment. Therefore, this paper presents a debate directed graph model for Weibo credibility evaluation. From the point of view of debate, the process of debate deduction is described visually and visually by the graph model. The topic corpus is used to form the dispute node, the subject emotion and the underlying logic relationship between the disputes are used to define the rule credibility, and the weights of the edges in the graph are set to represent the defense and attack intensity between the disputes. The credibility of the conclusion is obtained according to the correlation algorithm, and the credibility of the discriminant information is obtained by recursive evolution of the argument graph. The experimental results show that the synthetic index value of this model is 6 times higher than that of the traditional method.
【作者单位】： 昆明理工大学信息工程与自动化学院;云南省计算机技术应用重点实验室;昆明理工大学教育技术与网络中心;
【基金】：国家自然科学基金(81360230,81560296,61462056,61462051)
【分类号】：TP391.1;TP393.092
，

本文编号：1889805