Byte-Level Grammatical Error Correction Using Synthetic and Curated Corpora

TL;DR

This paper compares byte-level and subword-level models for grammatical error correction, demonstrating that byte-level models achieve higher accuracy across various error types, especially when trained on synthetic data and fine-tuned on real-world errors.

Contribution

It introduces a byte-level encoding approach for GEC and shows its superiority over subword models in correcting diverse error types, especially in morphologically rich languages.

Findings

Byte-level models outperform subword models in correction quality.

Synthetic training data combined with real error fine-tuning improves performance.

The approach is effective for morphologically rich languages like Icelandic.

Abstract

Grammatical error correction (GEC) is the task of correcting typos, spelling, punctuation and grammatical issues in text. Approaching the problem as a sequence-to-sequence task, we compare the use of a common subword unit vocabulary and byte-level encoding. Initial synthetic training data is created using an error-generating pipeline, and used for finetuning two subword-level models and one byte-level model. Models are then finetuned further on hand-corrected error corpora, including texts written by children, university students, dyslexic and second-language writers, and evaluated over different error types and origins. We show that a byte-level model enables higher correction quality than a subword approach, not only for simple spelling errors, but also for more complex semantic, stylistic and grammatical issues. In particular, initial training on synthetic corpora followed by finetuning on a relatively small parallel corpus of real-world errors helps the byte-level model correct a wide range of commonly occurring errors. Our experiments are run for the Icelandic language but should hold for other similar languages, particularly morphologically rich ones.