daikin-industries-ltd/ja-fineweb-2-hvac-high-quality-v1

本数据集是从ja-fineweb-2-hvac-fastText-scored-v1中提取的LLM评分4分及以上（高质量）的记录，专门针对空调（HVAC）相关的高质量文本数据。这些数据经过严格筛选，适合用于空调技术人员的教育。数据集包含文本内容、元数据（如URL、日期、语言等）、FastText分类结果和分数、LLM评分及其理由。数据格式为gzip压缩的JSONL格式。总记录数为22,587条，其中4分记录占62.5%，5分记录占37.5%。数据集使用MIT许可证，语言为日语。

This dataset is a subset of ja-fineweb-2-hvac-fastText-scored-v1, containing only records with LLM quality scores of 4 or higher (high-quality). It focuses on high-quality HVAC (Heating, Ventilation, and Air Conditioning) related text data, suitable for educating HVAC technicians. The dataset includes text content, metadata (such as URL, date, language, etc.), FastText classification results and scores, LLM scores, and rationales. The data is stored in gzip-compressed JSONL format. The total number of records is 22,587, with 62.5% scoring 4 and 37.5% scoring 5. The dataset is licensed under MIT and is in Japanese.