一种新的考虑注意力机制的微博谣言检测模型  

New Microblog Rumor Detection Model Based on Attention Mechanism

作　　者：潘德宇 宋玉蓉 宋波[2] PAN De-yu;SONG Yu-rong;SONG Bo(School of Automation and Artificial Intelligence,Nanjing University of Posts and Telecommunications,Nanjing 210023,China;School of Modern Post,Nanjing University of Posts and Telecommunications,Nanjing 210003,China)

机构地区：[1]南京邮电大学自动化学院、人工智能学院,南京210023 [2]南京邮电大学现代邮政学院,南京210003

出　　处：《小型微型计算机系统》2021年第2期348-353,共6页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61672298)资助;江苏高校哲学社会科学研究重点项目(2018SJZDI142)资助;教育部人文社会科学研究规划基金项目(17YJAZH071)资助.

摘　　要：随着社交媒体的迅速发展,谣言通过社交媒体迅速传播,识别社交媒体网络上的谣言是社交网络研究中一个至关重要的问题.本文提出了一种新的考虑注意力机制的微博谣言检测模型,考虑到卷积神经网络(CNN)提取到的特征对输出结果影响力问题,在经典的文本卷积神经网络(Text CNN)上引入了注意力机制,通过CNN中的卷积层学习微博窗口的特征表示,再根据每个特征表示对输出结果的影响力不同通过注意力机制赋予不同的权重来进行谣言事件的检测.研究结果表明,本文提出的微博谣言检测模型准确率达到了96.8%,并且在召回率和F1值上也有提升,即本文提出的新的微博谣言检测模型具有更好的谣言识别能力.With the rapid development of social media,rumors spread rapidly through social media,and identifying rumors on social media networks is a crucial issue in social network research.This paper proposes a new microblog rumor detection model that considers the attention mechanism.Considering the influence of the features extracted by the convolutional neural network(CNN)on the output result,it is introduced in the classic text CNN.In order to understand the attention mechanism,the feature representation of the microblog window is learned through the convolutional layer in CNN,and then according to the different influence of each feature representation on the output result,different weights are assigned to the attention mechanism to detect rumor events.The research results show that the accuracy rate of the microblog rumor detection model proposed in this paper reaches 96.8%,and the recall rate and F1 value are also improved,that is,the new microblog rumor detection model proposed in this paper has better rumor recognition ability.

关 键 词：谣言检测 卷积神经网络 注意力机制 准确率 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]