基于突发话题和领域专家的微博谣言检测方法

发布时间：2019-02-12 19:42

【摘要】：针对现有谣言检测方法中存在的数据采集困难和谣言检测滞后的问题,提出一种基于动量模型的突发话题检测和领域专家发现的谣言检测方法。该方法借鉴物理学中的动力学理论对话题特征进行建模,使用特征的动力学物理量描述特征的突发特性和发展趋势,并在对突发特征进行特征聚合之后提取得到突发话题;然后,依据话题与用户个人信息的领域相关性在候选专家池中发现领域相关的微博用户来甄别话题信息的真实性。基于新浪微博数据的实验结果表明,相对于仅基于有监督机器学习的微博谣言识别方法,该方法谣言识别准确率提高了13个百分点;相对于主流人工识别方法,将最长谣言检测用时缩短至20 h,能够较好地应用于实际的微博谣言检测环境。
[Abstract]:Aiming at the difficulty of data acquisition and the lag of rumor detection in the existing rumour detection methods, a novel method based on momentum model for burst topic detection and rumor detection by domain experts is proposed. The method uses dynamics theory in physics for reference to model the topic feature, uses the dynamic physical quantity of the feature to describe the burst characteristic and the development trend of the feature, and extracts the burst topic after the feature aggregation of the burst feature. Then, according to the domain relevance of topic and user's personal information, Weibo users are found in the pool of candidate experts to identify the authenticity of topic information. The experimental results based on Sina Weibo data show that the accuracy of this method is improved by 13 percentage points compared with the Weibo rumor recognition method based only on supervised machine learning. Compared with the mainstream manual recognition method, the longest rumour detection time can be shortened to 20 hours, which can be applied to the actual rumors detection environment of Weibo.
【作者单位】：四川大学计算机学院;四川大学网络空间安全学院;
【基金】：四川省教育厅重点资助项目(17ZA0238,17ZA0200) 四川省学术和技术带头人培养支持经费资助项目(2016)~~
【分类号】：G206;TP393.092

【相似文献】