DSTI/traffic-accidents-reports-5k

5k事故报告数据集包含大约5000行数据，旨在用于文本到文本生成的任务，即从5W1H（什么、何时、何地、谁、怎样、为什么、应急措施）信息生成一个段落式的事故报告。数据集分为训练集、验证集和测试集，分别包含大约4500、500和100条记录。数据集采用Parquet格式，以英文为主，适用于微调小型指令模型以生成简洁、结构化的事故报告。

The 5k Accident Reports dataset consists of approximately 5,000 rows of data designed for a text-to-text generation task, specifically converting 5W1H (What, When, Where, Who, How, Why, ContingencyActions) information into a single-paragraph accident report. The dataset is split into training, validation, and test sets with about 4,500, 500, and 100 examples respectively. The dataset is stored in Parquet format and is in English, intended for fine-tuning small instruction models to produce concise and structured incident reports.