To Err Is Human, but Llamas Can Learn It Too

arXiv:2403.05493v1 Announce Type: new
Abstract: This study explores enhancing grammatical error correction (GEC) through artificial error generation (AEG) using language models (LMs). Specifically, we fine-tune Llama 2-based LMs for error generation and find that this approach yields synthetic errors akin to human errors. Next, we train GEC Llama models with the help of these artificial errors and outperform previous state-of-the-art error correction models, with gains ranging between 0.8 and 6 F0.5 points across all tested languages (German, Ukrainian, and Estonian). …

abstract artificial arxiv cs.cl error error correction errors gec human language language models learn llama llama 2 llama models llamas lms next study synthetic through train type