APPLYING BERT MODEL ON CUSTOMER SENTIMENT ANALYSIS THROUGH COMMENTS

Tự Thanh Duy Nguyễn , Thanh Phước Trần , Thanh Trâm Trần , Quốc Tuấn Võ

Main Article Content

Abstract

In today's society, the development of e-commerce sites and social networks is growing strongly, along with e-commerce and social networks, it is certainly indispensable for comments expressing the people’s attitudes for a product or problem. Businesses always tried their best to capture the needs and attitudes of consumers with their products to the market. That is why in this study an application was built to analyze users’ attitudes through comments using the BERT model, comments were  collected on Shopee with Unilever brand. Besides comprising between PhoBERT and BERT with two other machine learning and deep learning models, KNN and LSTM were also used. In addition, the study also integrated some advanced technologies such as ReactJS for Frontend and FastAPI for Backend to deploy the application to a real website to increase the experience of a user. The initial results are very positive and can be applied to many other businesses.

Article Details

References

Devlin, J., Chang, M. W., Lee, K., & Toutanova, K. (2019). BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. In Proceedings of NAACL-HLT 2019, 4171-4186.
Guo, G., Wang, H., David, A. B.l, & Yaxin, B. (2004). KNN Model-Based Approach in Classification. On The Move to Meaningful Internet Systems 2003: CoopIS, DOA, and ODBASE, 986-996.
Nguyen, D. Q., & Nguyen, A. T. (2020). PhoBERT: Pre-trained language models for Vietnamese. Findings of the Association for Computational Linguistics: EMNLP 2020, 1037-1042.
Sepp, H., Jurgen S. (1997). Long Short-Term Memory. Neural Computation, 9(8), 1735-1780. https://doi.org/10.1162/neco.1997.9.8.1735