基于Attention与Bi-LSTM的谣言识别方法  

Rumor Recognition Method Based on Attention and Bi-LSTM

作　　者：冀源蕊 康海燕[1] 方铭浩 JI Yuanrui;KANG Haiyan;FANG Minghao(School of Information Management,Beijing Information Science and Technology University,Beijing 100192,China;Software R&D Center,Postal Savings Bank of China,Beijing 100160,China)

机构地区：[1]北京信息科技大学信息管理学院,北京100192 [2]中国邮政储蓄银行软件研发中心,北京100160

出　　处：《郑州大学学报（理学版）》2023年第4期16-22,共7页Journal of Zhengzhou University:Natural Science Edition

基　　金：国家社会科学基金项目(21BTQ079);教育部人文社会科学项目(20YJAZH046);国家自然科学基金项目(61370139)。

摘　　要：为了实现谣言的高效识别,提出一种基于Attention与Bi-LSTM的谣言识别方法。首先,设计一种基于双向循环神经网络的深度学习模型,并引入Attention机制对长序列编解码的时序问题进行关联,从每个句子中捕获最重要的语义信息,形成长期的记忆,从而高效识别谣言的二次传播;其次,设计Word Embedding机制,将文本数据映射到一个低维度的实数向量,避免了高维度的输入导致模型产生维度灾难;最后,在真实数据集上与先进的谣言识别方案进行对比,所提方法能达到94.3%的准确率,高于其他三种基于深度学习的谣言识别方案,从而验证了该方法的有效性。In order to achieve efficient rumor recognition,a rumor recognition method based on Attention and Bi-LSTM was proposed.Firstly,a deep learning model based on bidirectional recurrent neural network was designed,and the Attention mechanism was introduced to correlate the timing issues of long sequence encoding and decoding,so as to capture the most important semantic information from each sentence and form long-term memory,which could efficiently identify the secondary propagation of rumors.Secondly,the Word Embedding mechanism was designed to map the text data to a low-dimensional real number vector,so as to avoid the dimension disaster of the model caused by high-dimensional input.Finally,compared with advanced rumor recognition schemes on the real data sets,the accuracy of the proposed method could reach 94.3%,which was higher than the other three rumor identification schemes based on deep learning,thus the effectiveness of the method was verified.

关 键 词：谣言识别 辟谣机制 循环神经网络 Attention机制 

分 类 号：TP309.2[自动化与计算机技术—计算机系统结构]