Scholarlink

The emergence of large language models has opened promising opportunities for various natural language processing tasks, including sentiment analysis. This study investigates the capabilities of large language models in Vietnamese sentiment classification through both fine-tuning and prompting techniques. We evaluate models such as Bidirectional Encoder Representations from Transformers (BERT), Generative Pre-trained Transformer (GPT), LLaMA, and Mistral for this task. Experimental results highlight the superior performance of fine-tuning methods, with the Mistral 7B model achieving the highest F1-score on the UIT-VSFC student feedback dataset, outperforming advanced models such as TextGCN and BertGCN, as well as prompting-based approaches.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An Assessment of Large Language Models for Vietnamese Sentiment Classification

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Vu Thi Thanh Huong,
  • Thien Khai Tran

摘要

The emergence of large language models has opened promising opportunities for various natural language processing tasks, including sentiment analysis. This study investigates the capabilities of large language models in Vietnamese sentiment classification through both fine-tuning and prompting techniques. We evaluate models such as Bidirectional Encoder Representations from Transformers (BERT), Generative Pre-trained Transformer (GPT), LLaMA, and Mistral for this task. Experimental results highlight the superior performance of fine-tuning methods, with the Mistral 7B model achieving the highest F1-score on the UIT-VSFC student feedback dataset, outperforming advanced models such as TextGCN and BertGCN, as well as prompting-based approaches.