计算机技术与发展2024，Vol.34Issue(7)：175-183,9.DOI:10.20165/j.cnki.ISSN1673-629X.2024.0087
基于深度学习的微博疫情舆情文本情感分析
Sentiment Analysis of Weibo Epidemic Public Opinion Text Based on Deep Learning
摘要
Abstract
Public opinion sentiment analysis focuses on studying the public's emotional bias towards public events.Public opinions involving public health events will directly affect social stability,so sentiment analysis on Weibo is particularly important.We take text data sets related to the epidemic,and use RoBERTa,BiGRU,and the RoBERTa-BDA(RoBERTa-BiGRU-Double Attention)model combined with double-layer Attention as the overall structure.Firstly,RoBERTa is used to obtain word embedding representation of textual context information.Secondly,BiGRU is used to obtain the character representation,then the attention mechanism is used to calculate the global impact of each character,and then BiGRU is used to obtain the sentence representation.Finally,the Attention mechanism is used to calculate the weight ratio of each character to the sentence in which it is located,and the text representation of the full text is obtained,and the sentiment analysis is carried out through softmax function.In order to verify the effectiveness of the RoBERTa-BDA model,three experiments were designed.In the comparison experiment of different word vectors,the Macro F1 and Micro F1 values in RoBERTa compared with BERT increased by 0.42 percentage points and 0.84 percentage points,respectively,in different feature extraction layers.In the model comparison experiment,BiGRU-Double Attention increased by 3.62 percentage points and 1.34 percentage points compared to BiGRU-Attention.In the cross-platform comparison experiment,RoBERTa-BDA only decreased by1.29 percentage points and 2.88 percentage points on the Tieba platform Macro F1 and Micro F1 compared to the Weibo platform.关键词
RoBERTa/情感分析/特征提取/词向量/注意力机制/BiGRUKey words
RoBERTa/sentiment analysis/feature extraction/word vectors/attention mechanism/BiGRU分类
信息技术与安全科学引用本文复制引用
吴加辉,加云岗,王志晓,张九龙,闫文耀,高昂,车少鹏..基于深度学习的微博疫情舆情文本情感分析[J].计算机技术与发展,2024,34(7):175-183,9.基金项目
教育部人文社会科学研究青年基金(16YJCZH109) （16YJCZH109）
2022年陕西省科技计划项目之区域创新能力引导计划(2022QFY01-17) （2022QFY01-17）
智慧城市多模态场景感知关键技术研究以及应用(2023JH-RGZNGG-0011) （2023JH-RGZNGG-0011）
