全文下载：
202006014.pdf
文章编号： 1672-6987（2020）06-0099-08； DOI： 10.16351/j.1672-6987.2020.06.014
刘勇， 任晓伟(青岛科技大学 信息科学技术学院，山东 青岛 266061）
摘要： 随着电子商务的发展，用户评论文本呈爆炸式增长，对评论文本的情感倾向进行挖掘有助于为潜在用户及商家品牌提供观点倾向及决策支撑。传统文本情感分类方法存在深层语义挖掘不充分现象，导致下游分类任务效果不理想。本工作提出一种情感分类算法BERT-Bi-LSTM，基于BERT构建文本字向量表示，采用双向长短期记忆网络进行特征提取，结合注意力机制对用户评论文本进行积极、消极情感分类。本工作提出的方案在爬取京东评论数据集上准确率达到954%，在IMDB数据集上能达到9173%的准确率。实验结果表明，该模型有效提高了分类任务的准确率。
关键词： 文本情感分类； 词向量表示方法； 双向长短期记忆网络； 注意力机制
中图分类号： TP 391.1文献标志码： A
引用格式： 刘勇， 任晓伟. 一种基于深度学习的电商平台用户评论情感分类方法\[J\]. 青岛科技大学学报（自然科学版）， 2020， 41（6）： 99-106.
LIU Yong， REN Xiaowei. E-commerce platform user comment sentiment classification method based on deep learning\[J\]. Journal of Qingdao University of Science and Technology（Natural Science Edition), 2020， 41（6）： 99-106.
E
-
Commerce Platform User Comment Sentiment Classification
Method Based on Deep Learning
LIU Yong， REN Xiaowei
(College of Information Science and Technology, Qingdao University of Science and Technology, Qingdao 266061, China)
Abstract: With the development of e-commerce, user comment texts are exploding. Excavating the emotional tendency of the review text helps to provide perspectives and decision support for potential users and merchant brands. However, this method the traditional text sentiment classification tasks leads to insufficient deep semantic mining, and the downstream classification task is not ideal. This paper proposed a new method named BERT-Bi-LSTM. Through the pre-training model proposed by the Google team, applied to the sentence vector representation of Chinese text. The feature extraction is performed by bi-directional long short-term memory network（LSTM），and focus on key words through the attention mechanism. This method classifies Chinese comment texts as positive emotions and negative emotions. The proposed scheme in this paper has an accuracy rate of 954% on crawling the JD review data set, and an accuracy rate of 9173% on the IMDB dataset. Experimental results show that the model effectively improves the accuracy of classification tasks.
Key words: text sentiment classification; word vector representation; long short-term memory network; attention
收稿日期： 2019-10-23
基金项目： 山东省重点研发计划项目（GG201710030036）.
作者简介： 刘勇（1971—），女，博士，副教授.
