Leveraging LLMs for Bangla Grammar Error Correction:Error Categorization, Synthetic Data, and Model Evaluation
This work addresses the lack of resources for Bangla GEC, a domain-specific problem for Bangla speakers and NLP researchers, though it is incremental as it applies existing methods to a new language.
The paper tackled the underdeveloped problem of grammatical error correction (GEC) in Bangla by creating a synthetic dataset and instruction-tuning LLMs, resulting in a 3-7 percentage point improvement in performance compared to zero-shot settings and achieving human-like error identification.
Large Language Models (LLMs) perform exceedingly well in Natural Language Understanding (NLU) tasks for many languages including English. However, despite being the fifth most-spoken language globally, Grammatical Error Correction (GEC) in Bangla remains underdeveloped. In this work, we investigate how LLMs can be leveraged for improving Bangla GEC. For that, we first do an extensive categorization of 12 error classes in Bangla, and take a survey of native Bangla speakers to collect real-world errors. We next devise a rule-based noise injection method to create grammatically incorrect sentences corresponding to correct ones. The Vaiyakarana dataset, thus created, consists of 5,67,422 sentences of which 2,27,119 are erroneous. This dataset is then used to instruction-tune LLMs for the task of GEC in Bangla. Evaluations show that instruction-tuning with \name improves GEC performance of LLMs by 3-7 percentage points as compared to the zero-shot setting, and makes them achieve human-like performance in grammatical error identification. Humans, though, remain superior in error correction.