Issue |
ITM Web Conf.
Volume 70, 2025
2024 2nd International Conference on Data Science, Advanced Algorithm and Intelligent Computing (DAI 2024)
|
|
---|---|---|
Article Number | 04021 | |
Number of page(s) | 8 | |
Section | AI and Advanced Applications | |
DOI | https://doi.org/10.1051/itmconf/20257004021 | |
Published online | 23 January 2025 |
Sentiment Analysis of Product Reviews Using Fine-Tuned LLaMa-3 Model: Evaluation with Comprehensive Benchmark Metrics
School of Computer Science and Technology, China University of Mining and Technology, Nantong City, Jiangsu Province, 226000, China
Corresponding author: 03220957@cumt.edu.cn
Sentiment analysis, a crucial subfield of natural language processing, enables businesses and policymakers to understand public emotions and opinions, essential for crafting effective strategies across industries like marketing and customer service. As the volume of online reviews grows, automated sentiment classification models have become vital for efficiently processing this data. This study explores fine-tuning the LLaMA-8B large language model based on the Amazon Product Reviews dataset from Kaggle, aiming to improve sentiment classification accuracy. Using the LoRA fine-tuning approach combined with the Variant Greedy Search Technique (VGST) and TextBlob for polarity handling, the research addresses dataset size challenges. The model’s fine-tuning process includes one-shot learning and chain-of-thought prompting to better capture nuanced sentiment expressions. Evaluated using comprehensive metrics, LLaMA-8B demonstrates superior precision compared to Qwen2-7B and achieves near LLaVA performance with enhanced speed. Additionally, it outperforms models like Decision Tree, SVM, Multinomial NB, and XLNet in accuracy. This work underscores the potential of large language models for sentiment analysis and sets the stage for future extensions to multimodal input scenarios.
© The Authors, published by EDP Sciences, 2025
This is an Open Access article distributed under the terms of the Creative Commons Attribution License 4.0, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.