Aspect and Sentiment Detection in Vietnamese Text using Transfer Learning and LSTM

Abstract

Aspect detection and sentiment classification is a challenging task in the field of natural language processing (NLP), especially in Vietnamese due to its complex grammatical structure and diversity of expressions. In this study, we propose a method of using transfer learning (BERT) and LSTM to simultaneously detect aspects and sentiments on customer feedback on mobile phone products. BERT is used to extract semantic features from text, while LSTM processes these features to aspects detection and sentiment classification. These results show that our method outperforms traditional deep learning methods. The experiment results show that our model achieved an accuracy of 93.5%, F1-score of 88.7% on the overall performance of the model. For the aspects detections, our model get accuracy at 81.5% and F1-score at 89.9%. And for the sentiment classification, an accuracy of 90.6% and F1-score at 90.6% on all aspects of model. This paper compares the performance of BERT combined with RNN, GRU, and Transformer models, provides detailed evaluations of their effectiveness, discusses their limitations, and suggests future research directions to further enhance their practical applications.