JuDDGES/en-appealcourt-coded-instruct_v02

该数据集是从英格兰和威尔士刑事上诉法院的公开判决中收集的，包含了621个经过专家手工整理的判决，涵盖了46个不同的代码。每个例子都包含判决的全文、人类从判决中提取的JSON格式的值，以及人类编码到特定类别并规范化的值。该数据集旨在微调大型语言模型（LLMs）以可靠地从长篇法律文本中提取结构化数据，例如答辩日期、罪行等。数据集格式化为指令，适合信息提取和文本生成任务。

The dataset consists of 621 manually curated judgments from the Court of Appeal (Criminal Division) of England and Wales. Each example includes the full text of a judgment, human-extracted values from the judgment in JSON format, and human-coded values to specific categories, also in JSON format. The dataset is designed to fine-tune large language models (LLMs) for reliable extraction of structured data from lengthy legal texts, such as plea dates and offences. It is formatted as instructions, suitable for information extraction and text generation tasks.