融合自注意力机制的垃圾弹幕识别方法研究  

Research on Garbage Barrage Recognition Method Integrating Self Attention Mechanism

作　　者：费寅杰 黄旭 曾孟佳 Fei Yinjie;Huang Xu;Zeng Mengjia(School of Information Engineering,Huzhou University,Huzhou,China;School of Electronic Information,Huzhou College,Huzhou,China;Huzhou Key Laboratory for Urban Multidimensional Perception and Intelligent Computing,Huzhou,China)

机构地区：[1]湖州师范学院信息工程学院,浙江湖州 [2]湖州学院电子信息学院,浙江湖州 [3]湖州市城市多维感知与智能计算重点实验室,浙江湖州

出　　处：《科学技术创新》2024年第3期114-117,共4页Scientific and Technological Innovation

基　　金：浙江省软科学研究计划项目20YJCZH005。

摘　　要：网络直播的兴起,带来了各种类型的垃圾弹幕。传统的垃圾弹幕识别方法识别率低,无法满足目前直播平台垃圾弹幕识别需求。本文分析了垃圾弹幕文本特点,引入自注意力机制,结合ERNIE和TextCNN模型,设计了一种提高垃圾弹幕识别率的算法。通过对比实验,证明了此种方法的合理性和可行性,为今后相关领域的研究提供了一定的参考。The rise of online live streaming has brought various types of junk barrage.The traditional garbage barrage recognition method has a low recognition rate and cannot meet the current needs of garbage barrage recognition on live streaming platforms.This article analyzes the characteristics of garbage barrage text,introduces self attention mechanism,and combines ERNIE and TextCNN models to design an algorithm to improve the recognition rate of garbage barrage.Through comparative experiments,the rationality and feasibility of this method have been proven,providing a certain reference for future research in related fields.

关 键 词：短文本分类 直播弹幕 TextCNN模型 ERNIE模型 自注意力机制 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]