Geraldine/Ead-Instruct-full-175k

该数据集专为涉及编码档案描述（EAD）文件的XML格式的文本生成和完成任务设计。数据集提供了四种不同类型的提示-完成对，旨在帮助训练和评估语言模型在理解层次化XML结构、元素关系、基于路径的预测和掩码元素生成方面的能力。数据集来源于法国BnF收藏的`Inventaires du Catalogue général des manuscrits (CGM)`，并通过Python处理管道生成。数据集包含175,410个示例，分为完整数据集和两个子集。

This dataset is specifically designed for text generation and completion tasks involving Encoded Archival Description (EAD) files in XML format. The dataset provides four distinct types of prompt-completion pairs, each crafted to help train and evaluate language models in understanding hierarchical XML structures, element relationships, path-based predictions, and masked element generation. The dataset is derived from the French `Inventaires du Catalogue général des manuscrits (CGM)` of the BnF collections and was generated using a Python processing pipeline. It contains 175,410 examples, available as a full dataset and two subsets.