Editor/Reviewer login
Frequency : 12 issues per year
Subject : Computer Applications and Technology
ISSN : 2319–8656 (Online)
Home Archives Volume 12 Issue 6 ijcatr12061002

IJCATR Volume 12 Issue 6

An Aspect-Level Sentiment Analysis Approach Based on BERT and Attention Mechanism

Jinbo Liang, Hao Peng, Yuhao Zhan
10.7753/IJCATR1206.1002
keywords : BERT; sentiment classification; attention mechanism; global semantics; semantic interactive

PDF
In recent years, with the rapid increase in the number of comment texts on social media, more and more researchers have been studying the emotional tendencies in texts. In traditional sentiment analysis methods, there are still problems such as weak semantic dependency relationships and loss of semantic information caused by one-way networks. The paper proposes a sentiment classification method based on BERT and attention mechanism, which uses BERT as the word embedding model to obtain word vectors containing more semantic information, there by mitigating the impact of semantic sparsity ,the feature extraction layer uses bidirectional gated units to extract hidden vector information. The bidirectional network avoids the loss of forward semantics. The semantic interaction layer models the aspect words and context at the same time, and enhances the semantic dependency relationship between texts through interactive attention based on constructing global semantics. The experimental results show good performance.
@artical{j1262023ijcatr12061002,
Title = "An Aspect-Level Sentiment Analysis Approach Based on BERT and Attention Mechanism",
Journal ="International Journal of Computer Applications Technology and Research(IJCATR)",
Volume = "12",
Issue ="6",
Pages ="8 - 11",
Year = "2023",
Authors ="Jinbo Liang, Hao Peng, Yuhao Zhan"}
  • The paper proposes a sentiment classification method based on BERT and attention mechanism.
  • BERT as the word embedding model to obtain word vectors contain more semantic information.
  • The feature extraction layer uses bidirectional gated units to extract hidden vector information.
  • The semantic interaction layer enhances the semantic dependency relationship.
Copyright © 2012-2024 International Journal of Computer Applications Technology and Research All Rights Reserved.