基于BERT模型和双通道注意力的短文本情感分析方法  

Short Text Sentiment Analysis Method Based on BERT and Dual Channel Attention

作　　者：金华涛 JIN Huatao(Surfilter Network Technology Co.,Ltd.,Beijing 100029,China)

机构地区：[1]任子行网络技术股份有限公司,北京100029

出　　处：《信息与电脑》2021年第5期41-43,共3页Information & Computer

摘　　要：语篇情绪分析是自然语言处理的热门研究内容之一,已广泛应用在很多领域。因为主题建模能够发现隐含的语义结构,所以很多学者提出了基于主题模型的情感分析模型,虽然都考虑了语篇语境,但是还缺乏对情感语境的考虑和整合。针对情感分析任务和短文的特殊性,本文首先分析了BERT模型的发展历程和研究现状,进而分析了短文本情感分析网络设计,最后结合情感语境提出了基于词汇和主题的情感分类模型,希望能够为相关研究提供借鉴。Discourse sentiment analysis is one of the popular research contents of natural language processing,and it has been widely used in many fields.Because topic modeling can discover the implicit semantic structure,many scholars have proposed emotional analysis models based on topic models.Although they all consider the context of the text,they still lack consideration and integration of the emotional context.Aiming at the particularity of sentiment analysis tasks and short texts,this article first analyzes the development process and research status of the BERT model,and then analyzes the short text sentiment analysis network design,and finally proposes a vocabulary and topic-based sentiment classification model based on sentiment context,hoping to provide reference for related research.

关 键 词：BERT模型 双通道注意力 短文本情感分析 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]