jxtse/DSGram

数据集DSGram是为评估和发展大语言模型（LLMs）时代的语法错误校正（GEC）模型而设计的。它包含两个子集：DSGram-LLMs和DSGram-Eval。DSGram-LLMs是一个模拟数据集，包含由GPT-4注释的句子对，用于GEC模型的微调和成本效益评估。DSGram-Eval是一个手动注释的数据集，提供高质量的人类评分示例，以基准测试DSGram框架。数据集支持基于三个子指标的校正评估：语义连贯性、编辑级别和流利度。

The DSGram datasets are designed for the evaluation and development of Grammatical Error Correction (GEC) models in the era of large language models (LLMs). These datasets address key evaluation challenges by incorporating human annotations and LLM-generated scores. Two subsets are provided: DSGram-LLMs and DSGram-Eval. DSGram-LLMs is a simulated dataset containing GPT-4-annotated sentence pairs, enabling fine-tuning and cost-effective evaluation of GEC models. DSGram-Eval is a manually annotated dataset providing high-quality, human-scored examples to benchmark the DSGram framework. The datasets facilitate the evaluation of corrections based on three sub-metrics: Semantic Coherence, Edit Level, and Fluency.