metadata
dataset_info:
features:
- name: incorrect
dtype: string
- name: correct
dtype: string
splits:
- name: train
num_bytes: 193960468
num_examples: 540000
- name: test
num_bytes: 21546211
num_examples: 60000
download_size: 129134550
dataset_size: 215506679
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
language:
- ur
This dataset contains paired Urdu text for error correction tasks.
Data Format
Each dialogue entry in the dataset is structured as follows:
input_text: The text containing simulated errors.correct_text: The corresponding correct text.