融合TCN和BiLSTM的文本情感分析
Text sentiment analysis fusing TCN and BiLSTM
任楚岚 1仇全涛 1劣思敏1
作者信息
- 1. 沈阳化工大学计算机科学与技术学院,辽宁沈阳 110142;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
- 折叠
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.
Abstract
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 tempo-ral 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 indica-tors 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Key words
sentiment analysis/short text/mixed word embedding/deep learning/temporal convolutional network/bidirectional long short-term memory network/feed forward attention mechanism引用本文复制引用
基金项目
辽宁省教育厅科学研究基金项目(LJKZ0449)
辽宁省教育厅科学研究基金项目(LJKZ0434)
出版年
2024
NETL
NSTL
万方数据