You need to agree to share your Hugging Face information to access this dataset
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
You agree to not use the dataset to conduct experiments that cause harm to human subjects.
Log in or Sign Up to review the conditions and access this dataset content.
Roman Saraiki Parallel Dataset
Parallel English, Urdu, Roman Urdu and strict Jataki Saraiki data, in Shahmukhi and three Roman script styles, for low-resource translation, transliteration and language modeling.
About Saraiki
Saraiki (سرائیکی) is an Indo-Aryan language spoken primarily in Pakistan, particularly across southern Punjab and neighboring regions. It has a rich linguistic, literary, and cultural heritage, with its own vocabulary, grammatical structures, and regional varieties.
Saraiki is written in Shahmukhi (Perso-Arabic script), but in everyday digital communication (on phones, in chats and on social media) many speakers type it in Roman (Latin) script, the same way Urdu is typed as Roman Urdu.
Despite its significant speaker community, Saraiki remains a low-resource language in Natural Language Processing (NLP), and Roman Saraiki has almost no digital resources at all.
About This Dataset
The Roman Saraiki Parallel Dataset pairs every sentence across seven text columns: English, Urdu, Roman Urdu, Saraiki in Shahmukhi, and Saraiki in three Roman styles.
It is designed to support the development and evaluation of translation systems, transliteration models, Roman-script input understanding, and language technologies for Saraiki and Urdu.
Roman Saraiki Styles
Saraiki has five letters that Urdu does not. The three Roman columns differ only in how these are written; every other letter and the spelling of every word are identical across the three.
| Shahmukhi | roman_saraiki_sound |
roman_saraiki_typed |
roman_saraiki_casual |
|---|---|---|---|
| ٻ | ɓ | bb | b |
| ڄ | ʄ | jj | j |
| ݙ | ɗ | dd | d |
| ڳ | ɠ | gg | g |
| ݨ | ṇ | nr | n |
- sound keeps the exact pronunciation and maps one-to-one to Shahmukhi.
- typed keeps the sounds using only keyboard letters. Plain b, j, d, g are never doubled in this column, so a double letter always means a Saraiki-only sound.
- casual is how most people type Saraiki in chats; ٻال (child) and بال (hair) both become baal.
Fixed spellings used throughout the Roman Saraiki columns: kun (کوں), wich (وچ), te (تے), naal (نال), toon (توں), asaan (اساں), tusaan (تساں), naeen (نئیں), hai (ہے).
Categories and Lengths
17 categories: everyday, greetings, family, food, market, travel, health, farming, education, work, technology, chat, questions, time_numbers, feelings, culture, instructions.
| Length | Size | Share |
|---|---|---|
word |
1–3 words | 15% |
short |
4–10 words | 40% |
medium |
11–20 words | 30% |
long |
21–35 words | 15% |
Rows are balanced across every category × length combination.
Intended Uses
This dataset can support:
- Saraiki ↔ English, Saraiki ↔ Urdu and Urdu ↔ English machine translation
- Shahmukhi ↔ Roman Saraiki transliteration
- Urdu ↔ Roman Urdu transliteration
- Understanding Roman-script Saraiki and Urdu typed by real users
- Chat assistants and keyboards for Saraiki speakers
- NLP model training and fine-tuning
- LLM training and fine-tuning
- Low-resource language research
- Saraiki language technology
- Academic and educational projects
Why This Dataset Matters
Most Saraiki speakers who type on phones and social media write in Roman script, yet language technologies rarely support Roman Saraiki. Parallel data linking Roman Saraiki to Shahmukhi, Urdu, Roman Urdu and English is almost non-existent.
This dataset aims to contribute to the development of Saraiki translation, transliteration and Roman-script language technologies by providing aligned resources across scripts and languages.
Usage
from datasets import load_dataset
ds = load_dataset("themohal/roman-saraiki-parallel", split="train")
print(ds[0])
# e.g. Roman Saraiki (casual) -> Shahmukhi transliteration pairs
pairs = ds.select_columns(["roman_saraiki_casual", "saraiki_shahmukhi"])
There is a single train split. Make your own evaluation split if needed (for example with
ds.train_test_split(test_size=0.05, seed=42)).
Access Requirements
Access to this dataset is gated. Users are required to provide information about their intended use and agree to the applicable usage conditions before accessing the dataset.
By requesting access, users agree not to use the dataset to conduct experiments that cause harm to human subjects and agree to use the dataset for non-commercial purposes only.
Citation
If you use this dataset in research, publications, models, or publicly released projects, please cite and acknowledge the dataset:
@misc{raza2026romansaraiki,
title = {Roman Saraiki Parallel Dataset},
author = {Raza, Muhammad Farjad Ali},
year = {2026},
publisher = {Hugging Face},
howpublished = {\url{https://huggingface.co/datasets/themohal/roman-saraiki-parallel}}
}
Disclaimer
This dataset is provided for research and educational purposes. Users are responsible for ensuring that their use of the dataset complies with applicable laws, ethical requirements, and the dataset's license and access conditions.
- Downloads last month
- 92