【摘要】 为了更好地预测和理解公众情绪和舆情发展走向，本文分别利用卷积神经网络和长短期记忆网络捕捉文本的局部特征和全局特征，通过多头注意力机制加强模型对关键信息的关注，探究表情符号对文本分类的影响，并构建出对网络舆情进行高效处理的情感分析模型ECBL-MHA。实验结果表明，ECBL-MHA模型在对文本进行分类预测的准确率达到了90.51%，具备应用于情感分析的可行性。更多还原

【Abstract】 In order to better predict and understand the development trend of public emotions and public opinion,this paper uses CNN and BiLSTM to capture local and global features of text, strengthens the model’s attention to key information through multi head attention mechanism, explores the impact of emoticons on text classification,and constructs an efficient sentiment analysis model ECBL-MHA for online public opinion processing. The experimental results show that the ECBL-MHA model achieves an accuracy of 90.51% in text classification prediction, which is feasible for application in sentiment analysis.更多还原