README.md

metadata

dataset_info:
  features:
    - name: answer
      dtype: string
    - name: context
      dtype: string
    - name: question
      dtype: string
  splits:
    - name: train
      num_bytes: 5566439
      num_examples: 4997
  download_size: 3459636
  dataset_size: 5566439
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: mit
task_categories:
  - text-generation
  - text2text-generation
  - question-answering
language:
  - tr
pretty_name: Turkish Extractive Question-Answer Dataset
size_categories:
  - 1K<n<10K

Dataset Card for Dataset Name

The dataset consists of nearly 5000 {Context, Question, Answer} triplets in Turkish. It can be used in finetuning large language models for text-generation, masked language modeling, instruction following, and extractive question answering. The dataset is a manually curated version of multiple Turkish QA-based datasets and some of the answers are arranged by hand.