结合双向注意力机制的网络欺凌检测  

Combining with bidirectional attention mechanism for cyberbullying detection

作　　者：周杭霞 厉贤斌 崔晨 许瑞旭 ZHOU Hang-xia;LI Xian-bin;CUI Chen;XU Rui-xu(School of Information Engineering,China Jiliang University,Hangzhou 310018,China;Big Data and Network Security Research Institute,Zhejiang Police College,Hangzhou 310053,China;Software Development Department,Hanrun Ruihe(Zhejiang)Technology Co,Hangzhou 311100,China)

机构地区：[1]中国计量大学信息工程学院,浙江杭州310018 [2]浙江警察学院大数据与网络安全研究院,浙江杭州310053 [3]汉润瑞和(浙江)科技有限公司软件研发部,浙江杭州311100

出　　处：《计算机工程与设计》2025年第2期523-529,共7页Computer Engineering and Design

基　　金：公安部重点实验室开放课题基金项目(2021DSJSYS004);浙江省公益技术应用研究基金项目(LGG22E070003)。

摘　　要：针对网络欺凌文本内容嘈杂、文本特征交互不足的问题,提出一种结合双向注意力机制的网络欺凌检测模型。多尺度门控扩张因果卷积(MGDC)提取文本不同感受视野下的局部特征;双向门控循环单元(BiGRU)提取全局上下文语义特征;利用双向注意力机制学习全局上下文语义特征和局部特征之间的交互作用,弥补各自特征之间的不足。通过胶囊网络进行深层次的特征提取。通过实验验证了该方法在网络欺凌文本检测中的准确性和有效性。To address the problem of noisy text content and insufficient interaction of text features for cyberbullying,a cyberbullying detection model was proposed combining with bidirectional attention mechanism.Multi-scale gated dilated causal convolution(MGDC)was used to extract local features under different sensory horizons of the text.The bi-directional gated recurrent unit(BiGRU)was used to extract global contextual semantic features.The bidirectional attention mechanism was used to learn the interactions between the global contextual semantic features and the local features,to make up for the deficiencies of the respective features.The deep feature extraction was performed by the capsule network.The accuracy and effectiveness of the method in cyberbullying text detection are demonstrated through experiments.

关 键 词：网络欺凌 社交媒体 多尺度门控扩张因果卷积 双向注意力机制 胶囊网络 双向门控循环单元 特征提取 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]