2023
- FAIRSENTIMENT ANALYSIS FOR VIETNAMESE LANGUAGE USING PHOBERT MODEL2023
In modern days, it is critical and essential to extract information from unstructured text data in an automatic way. To improve the quality of provided services, many businesses and universities need to observe and analyze the feedback and evaluations from their customers and students. Therefore, Sentiment Analysis becomes the most crucial part of this improvement process. Many research articles that have proposed models for Vietnamese language with high accuracy, especially PhoBERT - a large-scale pre-trained language model for Vietnamese. In this study, we propose a PhoBERT model with some corrections in architecture. Additionally, we conducted some experiments trying to combine PhoBERT with some addition features from traditional methods such as TF-IDF and positive or negative features at the word level of the SentiWordNet dataset. Besides improving the model, during the research, we also observed and recognized some limitations in the UIT-VSFC dataset and suggested corrections to be more suitable. Our proposed model achieves state-of-the-art results on the original UIT-VSFC dataset with an F1-s ore of 94.28 % and an Accuracy of 94.5 %. With our proposed new version UIT-VSFC dataset, the model has significant results with an F1-score of 95.22 % and an Accuracy of 95.42 %.
@proceedings{sentiment, bibtex_show = {true}, abbr = {FAIR}, author = {Huynh, Thanh-Tu and Vo, Bach-Khoi and Ho, Anh-Dung and Vu, Duc-Lung}, pages = {447-454}, publisher = {Vietnam Academy of Science and Technology}, title = {SENTIMENT ANALYSIS FOR VIETNAMESE LANGUAGE USING PHOBERT MODEL}, year = {2023}, doi = {10.15625/vap.2022.0254}, organization = {Proceedings of the 15th National Conference on Fundamental and Applied Information Technology Research} }