Character Transformations for Non-Autoregressive GEC Tagging
This addresses GEC for morphologically rich languages, offering a more efficient alternative to word-based methods, though it is incremental as it builds on existing non-autoregressive GEC tagging approaches.
The paper tackles the problem of grammatical error correction (GEC) by proposing a character-based non-autoregressive approach with automatically generated character transformations, achieving solid results and dramatic speedup compared to autoregressive systems for Czech, German, and Russian.
We propose a character-based nonautoregressive GEC approach, with automatically generated character transformations. Recently, per-word classification of correction edits has proven an efficient, parallelizable alternative to current encoder-decoder GEC systems. We show that word replacement edits may be suboptimal and lead to explosion of rules for spelling, diacritization and errors in morphologically rich languages, and propose a method for generating character transformations from GEC corpus. Finally, we train character transformation models for Czech, German and Russian, reaching solid results and dramatic speedup compared to autoregressive systems. The source code is released at https://github.com/ufal/wnut2021_character_transformations_gec.