metadata
language:
- ur
size_categories:
- 100K<n<1M
task_categories:
- text-generation
- summarization
pretty_name: Urdu Instruct News Headline Generation
dataset_info:
features:
- name: inputs
dtype: string
- name: targets
dtype: string
- name: template_id
dtype: int64
- name: template_lang
sequence: string
splits:
- name: train
num_bytes: 255002720
num_examples: 100674
- name: test
num_bytes: 28284699
num_examples: 11187
download_size: 121546528
dataset_size: 283287419
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
Dataset Card for "Urdu-Instruct-News-Headline-Generation"
This Dataset is converted from the original dataset by Khalid Hussain, Nimra Mughal, Irfan Ali, Saif Hassan, Sher Muhammad Daudpota.
Task:
Generate the News Headline from the given News.
Split Size:
- train: 100674
- test: 11187
Prompt Template (In Urdu):
Random.choice b.w these 2. The first template is template_class 1, and the second template is template_class 2 in the dataset.
["اس اردو پیراگراف (خبروں) کا عنوان تجویز کریں
پیراگراف: {}",
"دیے گے خبروں کا عنوان تجویز کریں.
جملے: {}"
]
Translation:
1. Write a title for the following news article:
paragraph: {}
2. Suggest the title of the given the sentences
sentences: {}
Completion Template (In Urdu)
جی ضرور، یہ رہا آپ کے پیراگراف کا عنوان:
{}
Translation:
Sure, here is the title of the given article
{}