基于语义和情感协同感知的多域谣言检测  

Multi-domain Rumor Detection Based on Semantic and Emotional Co-perception

作　　者：刘伟 张明书[1] 魏彬 LIU Wei;ZHANG Ming-shu;WEI Bin(College of Cryptologic Engineering,Engineering University of PAP,Xi'an Shaanxi 710086,China;Academy of People's Armed Police,Beijing 100020,China)

机构地区：[1]武警工程大学密码工程学院,陕西西安710086 [2]武警部队研究院,北京100020

出　　处：《计算机仿真》2024年第12期544-549,共6页Computer Simulation

基　　金：国家社会科学基金(20BXW101)。

摘　　要：针对当前谣言检测模型大多数集中在单个领域方面且获取的语义特征与其内在情感融合度不够的问题,在仅使用谣言文本和所属领域信息的基础上,提出一种基于语义和情感协同感知的多域谣言检测模型,旨在利用协同注意力机制对谣言语义与其内在情感进行协同感知,以此模仿学习人类语言内含情感信息的行为特性。同时根据谣言所属领域的不同,基于自注意力机制进行自适应多域谣言检测。实验结果表明,上述模型在具有谣言领域标识的数据集上能够达到较好的效果,与单域模型RNN相比,在准确率和F1-Score上提高了6.62%,与现有多域模型相比,进一步提升了检测的性能。To address the problem that most of the current rumor detection models focus on a single domain and the semantic features obtained are not sufficiently integrated with their intrinsic emotions,this paper proposes a multidomain rumor detection model based on the synergistic perception of semantics and emotions,aiming at the synergistic perception of the semantics of rumors and their intrinsic emotions through a co-attentive mechanism to simulate learning the behavioral characteristics of human speech with intrinsic emotional information.It also performs adaptive multi-domain rumor detection based on the self-attention mechanism according to the domain to which the rumors belong.The experimental results show that the model can achieve better results on the dataset with rumor domain identification,improve the detection performance by 6.62%in terms of accuracy and F1-Score compared with the single-domain model RNN,and improve the detection performance compared with the existing multi-domain model.

关 键 词：谣言检测 协同注意力机制 域特征 语义特征 情感特征 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]