LLM Based Sentiment Classification From Bangladesh E-Commerce Reviews
2025-10-03
Summary
The study explores the use of large language models (LLMs) like Llama-3.1-8B for sentiment analysis of Bangla and English e-commerce reviews in Bangladesh. The fine-tuned Llama model outperformed other models, achieving an accuracy of 95.5% in sentiment classification. The research highlights effective parameter-efficient fine-tuning techniques like LoRA and PEFT for handling mixed-language texts in low-resource settings.
Why This Matters
Understanding consumer sentiment is crucial for e-commerce platforms, especially in linguistically diverse markets like Bangladesh. This study demonstrates that advanced LLMs can effectively analyze sentiment in mixed-language reviews, offering businesses valuable insights into customer opinions. The research also highlights the potential of LLMs to overcome challenges in low-resource languages, paving the way for broader applications in multilingual sentiment analysis.
How You Can Use This Info
Professionals in e-commerce and marketing can leverage LLMs to gain deeper insights from customer reviews, enabling more informed decision-making and improved customer engagement strategies. By adopting parameter-efficient models like Llama-3.1-8B, businesses can efficiently process customer feedback even with limited computational resources. Additionally, this approach can be applied to other low-resource languages, expanding the scope of sentiment analysis in global markets.