基于BERT模型的短视频平台评论信息情感分析研究  

Research on Sentiment Analysis in Comments of Short Video Platform Based on BERT Model

作　　者：程顺正 胡楠[1] CHENG Shunzheng;HU Nan(School of Electronic and Information Engineering,Liaoning Institute of Science and Technology,Benxi Liaoning 117004,China)

机构地区：[1]辽宁科技学院电子与信息工程学院,辽宁本溪117004

出　　处：《辽宁科技学院学报》2024年第6期30-33,共4页Journal of Liaoning Institute of Science and Technology

基　　金：辽宁省教育厅高等学校基本科研项目(2024JYTYB-04)。

摘　　要：为深入挖掘社交媒体中的用户生成内容(UGC)的丰富情感信息,文章提出了一种基于BERT(Bidirectional Encoder Representations from Transformers)模型的短视频平台评论情感分析方法。文章对短视频平台的评论数据进行了收集和预处理,包括去除噪声、分词和标注情感标签,利用BERT模型对处理后的评论进行特征提取,利用其上下文理解能力,以捕捉评论中的细微情感变化,通过构建情感分类模型和迁移学习的方式对BERT进行微调。实验结果表明,基于BERT的情感分析模型在短视频评论情感分类任务中表现优异,相较于传统的情感分析方法,准确率和F1值均有显著提升,准确识别用户对视频内容的态度,可帮助内容创作者和平台更好地分析市场需求。To deeply explore the rich emotional information of user generated content(UGC)in social media,this paper proposes a sentiment analysis method for short video platform comments based on the BERT(Bidirectional Encoder Representation from Transformers)model.The paper collects and preprocesses comment data from short video platforms,including removing noise,seg⁃menting words,and annotating emotional tags.The BERT model is used to extract features from the processed comments,and its contextual understanding ability is utilized to capture subtle emotional changes in the comments.An emotion classification model is developed,and the BERT model is fine-tuned through transfer learning.The experimental results show that the sentiment analysis model based on BERT performs well in the task of sentiment classification of short video comments.Compared with traditional sen⁃timent analysis methods,the accuracy and F1 score are significantly improved,which can identify users'attitudes towards video content and help content creators and platforms better analyze their market demands.

关 键 词：BERT模型 情感分析 二分类 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]