文献类型：会议论文

英文题名：Sentiment Analysis of Text Based on CNN and Bi-directional LSTM Model

作者：Zhou, Kai Long, Fei

第一作者：Zhou, Kai

通信作者：Long, F[1]|[144401cdcd7b1b0915448]龙飞;

机构：[1]Guizhou Univ, Coll Big Data & Informat Engn, Guiyang, Guizhou, Peoples R China;[2]Guizhou Inst Technol, Sch Elect Engn, Guiyang, Guizhou, Peoples R China

第一机构：Guizhou Univ, Coll Big Data & Informat Engn, Guiyang, Guizhou, Peoples R China

通信机构：corresponding author), Guizhou Inst Technol, Sch Elect Engn, Guiyang, Guizhou, Peoples R China.|贵州理工学院;

会议论文集：24th IEEE International Conference on Automation and Computing (ICAC) - Improving Productivity through Automation and Computing Newcastle

会议日期：SEP 06-07, 2018

会议地点：Newcastle Univ, Newcastle upon Tyne, ENGLAND

主办单位：Newcastle Univ

语种：英文

外文关键词：sentiment analysis; Chinese reviews text; CNN; Bi LSTM

年份：2018

摘要：In order to overcome the deficiency of sentiment analysis based on traditional machine learning, which difficulty of effective feature selection and inadequacy of marked training corpus will affect the performance of the classification system, we address the sentiment emotions analysis problem of Chinese product reviews text by combining convolutional neural network (CNN) with bidirectional long-short term memory network (BiLSTM) in this paper. The CNN can extract the sequence features from the global information, and it is able to consider the relationship among these features. The BiLSTM not only solves the long-term dependency problem, but also considers the context of the text at the same time. The result of numerical experiments shows that the proposed model achieves better metrics performance than the state-of-the-art methods.