Leveraging LLMs for Bangla Grammar Error Correction:Error Categorization, Synthetic Data, and Model Evaluation

TL;DR

This paper explores leveraging large language models for Bangla grammatical error correction by categorizing errors, creating a synthetic dataset, and evaluating model performance, achieving near-human accuracy in error detection.

Contribution

It introduces a comprehensive error categorization, a large synthetic dataset, and demonstrates improved LLM performance in Bangla GEC through instruction tuning.

Findings

Instruction-tuned LLMs improve GEC accuracy by 3-7 percentage points.

The Vaiyakarana dataset contains over 5.6 lakh sentences with errors.

Humans still outperform LLMs in error correction tasks.

Abstract

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.