摘要
微博文本情绪分析技术在舆情监控等领域具有广泛应用。基于传统机器学习模型和情感词典进行情感分析的结果往往不够理想,如何提升性能成为该领域的一个主要挑战。本文中我们使用了基于深度学习的BERT以完成语言理解任务并与传统做法性能相比较,结果中BERT模型取得了更好的性能。之后我们利用该模型进行三分类以分析COVID-19疫情期间的微博评论,总体上正面与中立情绪占主导。此外,我们也针对词频和词云进行相关分析,以期实现全方面了解此次疫情期间社会情感状态的目的。
Sentiment analysis of microblog text is widely used in public opinion monitoring and other fields.The results of sentiment analysis based on traditional machine learning models and sentiment dictionaries are often not ideal.How to improve performance has become a major challenge in this field.In this thesis,we use BERT based on deep learning to complete the language understanding task.Compared with traditional methods,BERT model has achieved better performance.We use the model to analyze microblog comments during the COVID-19 epidemic by conducting a three-category classification and find that positive and neutral emotions are dominant.We also conduct further analysis on word frequency and word cloud to gain more insights into the emotional states during the epidemic.
作者
刘洪浩
LIU Hong-hao(College of International Education,Henan University,Kaifeng 475000,China)
出处
《软件》
2020年第9期185-188,共4页
Software
