CoolFace
Datasetpublic

proxectonos/nos-rag-news

NOS RAG Dataset (Galician News) Dataset Description This dataset is built around a collection of Galician news articles and a set of question–answer pairs derived from them. The main goal is to provide a compact benchmark for evaluating Retrieval-Augmented Generation (RAG) systems in a realistic setting. Each question is grounded in a specific news article, and answering it correctly requires identifying and using the relevant parts of the source text. In many… See the full description on the dataset page: https://huggingface.co/datasets/proxectonos/nos-rag-news.

sourceHugging Faceapache-2.0updated 5mo agoView on Hugging Face
0likes22downloads
Dataset Card

NOS RAG Dataset (Galician News)

Dataset Description

This dataset is built around a collection of Galician news articles and a set of question–answer pairs derived from them. The main goal is to provide a compact benchmark for evaluating Retrieval-Augmented Generation (RAG) systems in a realistic setting.

Each question is grounded in a specific news article, and answering it correctly requires identifying and using the relevant parts of the source text. In many cases, the dataset includes multiple answer variants and references to specific paragraphs, which makes it suitable for testing both retrieval quality and answer generation.

Although it is primarily intended for RAG evaluation, the dataset can also be used for more traditional question answering (QA) tasks, including both extractive and generative approaches.

Dataset Structure

The dataset is split into two parts:

1. Questions (questions config)

This file contains the QA pairs linked to the source documents.

Each example includes:

  • —id: Unique identifier
  • —source_id: Identifier of the source news article
  • —question: A question in Galician
  • —answer: One or more valid answers
  • —context: Text passage(s) supporting the answer
  • —category: Topic of the question (e.g., politics, economy, technology)
  • —context_paragraph_indices: Indices pointing to relevant paragraphs in the source document

A single article can be associated with multiple questions.

2. News Articles (news config)

This file contains the original news articles used to generate the QA pairs.

Each entry corresponds to a full document and includes:

  • —id (int): A unique identifier for the document within the dataset.
  • —source (string): Name of the original media outlet or source where the article was published.
  • —source_id (string): Identifier of the document within its source. This field is used to link the document with the questions split, where it appears as the source_id of each QA pair.
  • —title (string): Title of the news article.
  • —text (list[string]): The full content of the article, split into paragraphs. Each element in the list corresponds to a paragraph, which enables fine-grained retrieval and aligns with the contextparagraphindices field in the questions split.

Language of the document (e.g., gl for Galician). While the dataset is monolingual, this field allows for easy filtering or future extension to multilingual settings.

These documents provide the full context needed to simulate a real retrieval scenario, where relevant information must be located within longer texts rather than pre-selected snippets.

Loading the Dataset

You can load the dataset using the 🤗 datasets library.

Load questions

python
from datasets import load_dataset

dataset = load_dataset("proxectonos/nos-rag-news", "questions")

print(dataset["questions"][0])

Load news articles

python
from datasets import load_dataset

dataset = load_dataset("proxectonos/nos-rag-news", "news")

print(dataset["news"][0])

Licensing

This dataset is released under the Apache 2.0 license. Please ensure that usage of the original news content complies with their respective terms.

Acknowledgements

This work is funded by the Ministerio para la Transformación Digital y de la Función Pública - Funded by EU – NextGenerationEU within the framework of the project Desarrollo de Modelos ALIA. (Esta publicación del proyecto Desarrollo de Modelos ALIA está financiada por el Ministerio para la Transformación Digital y de la Función Pública y por el Plan de Recuperación, Transformación y Resiliencia – Financiado por la Unión Europea – NextGenerationEU)