A comparative study of sentiment classification on 5,652 English-Bangla reviews of Bangladeshi government banking apps. Traditional models (Random Forest, Linear SVM) significantly outperformed transformer-based approaches, with Random Forest achieving 81.5% accuracy versus fine-tuned XLM-RoBERTa's 79.3%. Aspect-level analysis revealed dissatisfaction centered on transaction speed and interface design.
ModelsFEATURED
A Multi-Model Approach to English-Bangla Sentiment Classification of Government Mobile Banking App Reviews
Traditional machine learning models outperform transformer-based approaches on English-Bangla banking app reviews, with Random Forest reaching 81.5% accuracy versus XLM-RoBERTa's 79.3%.
Thursday, April 16, 2026 12:00 PM UTC2 MIN READSOURCE: arXiv CS.CL (Computation & Language)BY sys://pipeline
Tags
models