proxectonos/dog-rag
收藏资源简介:
该数据集包含从西班牙加利西亚自治区官方公报(Diario Oficial de Galicia, DOG)的出版物中提取的问题-答案三元组。每个示例基于真实的法律和行政文档,涵盖公共就业、住房补贴、文化遗产和机构法规等多个领域。数据集主要用于评估检索增强生成(RAG)系统,每个示例包含一个或多个上下文段落以及释义的问题和答案,可用于测试系统是否能正确检索信息并生成响应。此外,它也适用于标准的问答任务(包括抽取式和生成式)。所有数据均为加利西亚语,因此对多语言或低资源自然语言处理研究具有价值。数据集包括三个配置:triplets(默认QA三元组)、dog(简化版文档数据)和dog_metadata(带元数据的文档数据),每个条目包含唯一标识符、源文档信息、URL、问题列表、答案列表、上下文列表和类别字段。
This dataset contains question–answer triplets derived from publications of the Diario Oficial de Galicia (DOG), the official gazette of the autonomous community of Galicia (Spain). Each example is grounded in real legal and administrative documents, covering a wide range of domains such as public employment, housing subsidies, cultural heritage, and institutional regulations. The dataset is primarily intended for evaluating Retrieval-Augmented Generation (RAG) systems, with each example including one or more context passages along with paraphrased questions and answers, making it useful for testing retrieval and generation capabilities. It can also be used for standard question answering (QA) tasks, both extractive and generative. All data is in Galician, making it relevant for multilingual or lower-resource NLP settings. The dataset includes three configurations: triplets (default QA triplets), dog (simplified document data), and dog_metadata (document data with metadata), with each entry containing fields such as uid, file_name, url, question list, answer list, context list, and category.




