Refining Czech GEC: Insights from a Multi-Experiment Approach

2025-06-27

AI Summary

To address the suboptimal performance and efficiency of Czech grammatical error correction (GEC), this paper proposes a Transformer-based neural machine translation framework. Its core innovation is a dynamic hybrid synthetic error generation pipeline that jointly incorporates language-agnostic error patterns and Czech-specific linguistic rules, enhanced by domain-balanced sampling and fine-grained subword tokenization for efficient data augmentation. We systematically investigate the impacts of corpus selection, error injection strategies, and model scale, and evaluate large-model adaptation under both user-prompt fine-tuning and expert-annotated fine-tuning paradigms. Experiments demonstrate state-of-the-art results on the CzechGEC benchmark, with substantial improvements in correction accuracy and faster inference speed compared to existing approaches. The trained models and source code are publicly released.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Calibration & Uncertainty Quantification
• Search and Optimization: Learning to Search

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

We present a grammar error correction (GEC) system that achieves state of the art for the Czech language. Our system is based on a neural network translation approach with the Transformer architecture, and its key feature is its real-time synthetic generation pipeline, which dynamically augments sentences with artificial errors by introducing both language-agnostic and Czech-specific errors. We conduct a comprehensive series of experiments, investigating the Czech GEC corpora as bases for synthetic error introduction, several error generation strategies, domain balancing, tokenization granularity, model size, and data scaling during fine-tuning. Additionally, we evaluate the performance of large language models (LLMs) on Czech GEC in both end-user and expert fine-tuning scenarios. Our best-performing model is superior both in performance and computational efficiency. The source code and the trained model links are available on https://github.com/ufal/tsd2025-gec.

Problem

• Develop state-of-the-art Czech grammar error correction system
• Explore synthetic error generation for Czech-specific GEC training
• Evaluate LLMs and optimize model efficiency for Czech GEC

Innovation

• Transformer-based neural network for Czech GEC
• Real-time synthetic error generation pipeline
• Comprehensive multi-experiment optimization approach