融合TCN和BiLSTM的文本情感分析  

Text sentiment analysis fusing TCN and BiLSTM

作　　者：任楚岚 仇全涛 劣思敏 REN Chu-lan;QIU Quan-tao;LIE Si-min(College of Computer Science and Technology,Shenyang University of Chemical Technology,Shenyang 110142,China;Liaoning Key Laboratory of Intelligent Technology for Chemical Process Industry,Shenyang University of Chemical Technology,Shenyang 110142,China)

机构地区：[1]沈阳化工大学计算机科学与技术学院,辽宁沈阳110142 [2]沈阳化工大学辽宁省化工过程工业智能化技术重点实验室,辽宁沈阳110142

出　　处：《计算机工程与设计》2024年第7期2090-2096,共7页Computer Engineering and Design

基　　金：辽宁省教育厅科学研究基金项目(LJKZ0449、LJKZ0434)。

摘　　要：为在短文本语义情感分析过程中对词嵌入对情感语义充分表达,挖掘深层次语义信息,提出一种融合TCN和改进BiLSTM的文本情感分析算法。通过混合词嵌入对短文本向量化;将训练后的词向量先输入时序卷积网络,后输入到改进的双向长短时记忆网络中提取情感特征;强制向前注意力机制对提取到的特征进行加权;通过softmax函数进行情感分类输出。通过在新冠疫情评论数据集建模,模型的各项指标均达到92%以上,相较于其它模型性能更优。To fully express the emotional semantics of word embedding in the process of semantic sentiment analysis of short texts,and to mine deep semantic information,a text sentiment analysis algorithm integrating TCN and improved BiLSTM was proposed.The short text was vectorized through mixed word embedding.The trained word vector was first input into the temporal convolutional network and then input into the improved bidirectional long-short-term memory network to extract emotional features.The forced forward attention mechanism was used to weight extracted emotional features.The emotional classification output was made through the softmax function.Through modeling on the new crown epidemic comment dataset,all the indicators of the model reach more than 92%,which are better than that of other models.

关 键 词：情感分析 短文本 混合词嵌入 深度学习 时序卷积网络 双向长短时记忆网络 强制向前注意力机制 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]