基于AM的BLSTM网络电商评论情感倾向分类

Sentiment classification of online retailer reviews based on AM and BLSTM networks

针对电商评论中所包含的消费者情感倾向信息问题,提出一种基于注意力机制和双向长短期记忆(bidirectional long-short term memory,BLSTM)网络的情感倾向分类模型。该模型使用预训练的字向量作为输入特征,通过双向长短期记忆网络来学习文本的语义特征。依此特征,设计了一种新的注意力机制来捕捉BLSTM模型生成的文本语义特征中重要的信息,以降低文本中冗余噪声对于情感倾向分类的影响。实验结果表明,与传统机器学习方法以及长短期记忆模型和双向长短期记忆模型相比,所提出模型在电商评论的情感倾向分类上取得了较好的结果。

Aiming at the problem of consumer sentiment information contained in e-commerce reviews,a sentiment classification model based on attention model and bidirectional long short term memory(BLSTM)network is proposed.The model uses the pre-trained word vector as the input feature,and learns the semantic features of the text through the two-way short-term memory network.On this basis,a new attention mechanism is designed to capture the important information in the text semantic features generated by blstm model,so as to reduce the impact of redundant noise on sentiment classification.The experimental results show that compared with the traditional machine learning method,the short-term memory model and the two-way short-term memory model,the proposed model has achieved better results in the classification of e-commerce comments' emotional tendencies。