基于BLSTM和注意力机制的电商评论情感分类模型

Sentiment Classification Model Based on BLSTM and Attentional Mechanism

随着互联网的飞速发展,电商评论中所包含的情感信息对商家愈发重要。面对海量数据,传统的基于情感词典和机器学习算法来进行情感分类的方法已经不再适用。为了有效学习文本特征,减少文本中冗余噪声对于情感分类的影响,提出一种基于双向长短时记忆网络(BLSTM)和注意力机制的情感分类模型。实验表明,相较于传统的机器学习方法和普通的深度学习方法,论文模型在准确率、召回率和F1值指标上均有明显提高。

With the rapid development of the Internet,the sentiments contained in ecommerce reviews is increasingly impor. tant to businesses. In the face of massive data,traditional methods of sentiment classification based on sentiment dictionaries and machine learning methods are no longer suitable for use. In order to effectively learn the text features and reduce the impact of redun. dant noise during the sentiment classification,a sentiment classification model based on bi-directional short and long-term memory network(BLSTM)and attentional mechanism is proposed. The experiment results show that compared with the traditional machine learning methods and ordinary deep learning methods,the model in the precison,recall rate and F1 score have significantly im. proved.