A Comparative Study of Sequential Models and Transformer-Based LLMs for Bangla Suspicious Political Comment Classification

DOI

10.2991/978-94-6239-664-7_32How to use a DOI?

Keywords

Bangla NLP; Suspicious Comment Detection; Political Text Classification; BanglaBERT; LSTM; Bi-LSTM; Transformer Models; Low-Resource Languages; Deep Learning; Contextual Embeddings

Abstract

The rapid worldwide spread of political discussions in Bangla calls for the automation of detection of suspicious or harmful content. This paper considers suspicious political comments in Bangla classification with both sequential-based and transformer-based deep learning names. The dataset is the Suspicious Bangla Text Dataset, which consists of 43,389 comments labeled as suspicious or non-suspicious. The dataset has undergone rigorous preprocessing: normalization, tokenization, and sequence padding. The two Recurrent Neural Networks, LSTM and Bi-LSTM, and two transformer models, BanglaBERT and mBERT, were trained with stratified 80–20 splits and tested. Experimentally, the results show that transformer-based models, especially BanglaBERT, got better than the sequential network by getting 93% overall accuracy, followed by the balanced precision, recall, and F1-score for both classes. These recent developments show how they were able to do contextual embedding and fine-tuning in low-resource languages by setting a benchmark framework with which suspicious political content in the Bangla language can be detected for safer online discourse and future NLP research in low-resource.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Rasel Parvez
AU  - Md. Ikramul Hossain
AU  - Sadman Sadik Khan
AU  - S. M. Aminul Haque
AU  - Sami Ahmed
AU  - Abu Hurairah Rifat
PY  - 2026
DA  - 2026/06/08
TI  - A Comparative Study of Sequential Models and Transformer-Based LLMs for Bangla Suspicious Political Comment Classification
BT  - Proceedings of the International Conference on Intelligent Data Analysis and Applications (IDAA 2025)
PB  - Atlantis Press
SP  - 459
EP  - 472
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6239-664-7_32
DO  - 10.2991/978-94-6239-664-7_32
ID  - Parvez2026
ER  -