Publications
Research on low-resource Bangla NLP, mostly datasets and methods that move the state of the art when labelled data is scarce.
- EMNLP 2023
BanLemma: A Word Formation Dependent Rule and Dictionary Based Bangla Lemmatizer
Findings of the Association for Computational Linguistics, EMNLP 2023
- Rule-based Bangla lemmatizer built on iterative suffix stripping conditioned on the part-of-speech tag of the word.
- Outperforms every previously published Bangla lemmatisation method on existing datasets.
Supervised by Dr. Nabeel Mohammed, Dr. Ruhul Amin
- RANLP 2023
BanglaBait: Semi-Supervised Adversarial Approach for Clickbait Detection on Bangla Clickbait Dataset
Recent Advances in Natural Language Processing, RANLP 2023
- First Bangla clickbait news dataset, with 15,056 labelled articles.
- Shows that semi-supervised adversarial training beats supervised baselines on this task.
Supervised by Dr. Farig Sadeque
- DCAI 2022
A GAN-BERT Based Approach for Bengali Text Classification with a Few Labeled Examples
Distributed Computing and Artificial Intelligence (DCAI 2022), Springer
- Trains transformer classifiers adversarially with a GAN so they work when very few labelled examples exist.
- First paper to apply GAN-BERT to Bangla text classification.
Supervised by Annajiat Alim Rasel