Skip to content

Dataset · Metin

Turkish question-answer pairs for Wikipedia RAG

Metin/WikiRAG-TR

Synthetic dataset of nearly 6,000 Turkish question-answer pairs for retrieval-augmented generation systems.

What's inside

The dataset contains 5,999 Turkish examples with a total size of 20.5 MB: 5,725 synthetically generated question-answer pairs and 274 negative examples. The materials are created based on the introductory sections of Turkish Wikipedia articles. Data is presented in text and tabular formats under the Apache 2.0 license.

What you can do with it

  • Training and evaluating Turkish RAG systems
  • Fine-tuning models for context-based question answering
  • Training rankers using negative examples
Tasks
Question answering
Languages
Turkish
Size
1K–10K rows
Downloads this month
380
Likes
37

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("Metin/WikiRAG-TR")