RoLegalGEC
收藏资源简介:
RoLegalGEC是由布加勒斯特理工大学等机构联合推出的首个罗马尼亚语法律领域语法错误检测与修正数据集,包含35万条法律文本段落及其错误标注。数据集通过合成生成方法构建,结合语法规则逆向应用、噪声注入及大语言模型提示技术,覆盖20类罗马尼亚语典型错误。该资源旨在支持法律文档自动校对系统的开发,填补低资源语言在专业领域语法纠错研究的空白。
RoLegalGEC is the first Romanian legal-domain grammatical error detection and correction dataset, jointly developed by institutions including the Polytechnic University of Bucharest and other relevant partners. It contains 350,000 legal text paragraphs with corresponding error annotations. The dataset is constructed via synthetic generation methods, which integrate reverse application of grammatical rules, noise injection, and large language model (LLM) prompting techniques, covering 20 typical types of Romanian grammatical errors. This resource aims to support the development of automatic legal document proofreading systems, and fills the research gap in professional-domain grammatical error correction for low-resource languages.




