Scopus Indexed Publications

Paper Details


Title
Explainable Sentiment Analysis of E-Commerce Product Reviews in Bangla Using Attention-Based BiLSTM

Author
M. Humayet Islam, Abdullah Raian, Faria Afsana Rimu, Neha Khan, Rejowan Arifin Nayeem, Sumiya Rahman,

Email

Abstract

In the age of digitization, customer reviews on ecommerce websites contain valuable insights on consumer satisfaction, product quality, and effectiveness of services. Sentiment analysis, a subfield of natural language processing (NLP), enables automatic classification of reviews to determine public opinion. In this work, a sentiment analysis system is presented for e-commerce product reviews in Bangla using both traditional machine learning models and deep learning approaches. A manually labeled dataset of 1,000 reviews and an augmented dataset of 5,000 reviews were used for training and evaluation. Word embeddings were applied for feature representation, and Explainable AI (XAI) with attention scores was used to identify key tokens influencing predictions. Among all models, BiLSTM with Attention achieved the best performance, with 94.00% accuracy on the original dataset and 98.80% on the augmented dataset. This study highlights the effectiveness of combining augmentation, deep learning, and attention mechanisms in enhancing automatic review analysis for e-commerce platforms.


Keywords

Journal or Conference Name
2025 IEEE 4th International Conference on Robotics, Automation, Artificial-Intelligence and Internet-of-Things, RAAICON 2025

Publication Year
2025

Indexing
scopus