nassimjp/Polyglot-Restaurant-Questions-Dataset

🍽️ Polyglot Restaurant Questions Dataset

A multilingual dataset of 101 restaurant / catering order cancellation questions translated into 6 languages: English, Farsi (Persian), Japanese, Pashto, Sindhi, and Urdu.

This dataset is designed for intent classification, customer-support chatbots, question-answering systems, and cross-lingual NLP research in the food-service domain.

🌍 Languages

Total records: 606

📦 Data Format

Each file is in JSONL format with the following schema:

{"id": "0", "question": "My plans have changed, how can I cancel my catering order?"}

🚀 Usage

Load a single language

from datasets import load_dataset

ds = load_dataset(
    "nassimjp/Polyglot-Restaurant-Questions-Dataset",
    "urdu",
)
print(ds["train"][0])
# {'id': '0', 'question': 'میرے منصوبے بدل گئے ہیں، میں اپنا کیٹرنگ آرڈر کیسے منسوخ کر سکتا ہوں؟'}

Load all languages at once

from datasets import load_dataset

ds = load_dataset(
    "nassimjp/Polyglot-Restaurant-Questions-Dataset",
    data_files={
        "english":  "data/english.jsonl",
        "farsi":    "data/farsi.jsonl",
        "japanese": "data/japanese.jsonl",
        "pashto":   "data/pashto.jsonl",
        "sindhi":   "data/sindhi.jsonl",
        "urdu":     "data/urdu.jsonl",
    },
)
print(ds)

Load with huggingface_hub

from huggingface_hub import hf_hub_download

path = hf_hub_download(
    repo_id="nassimjp/Polyglot-Restaurant-Questions-Dataset",
    filename="data/urdu.jsonl",
    repo_type="dataset",
)
print(path)

🎯 Intended Use Cases

• —Intent classification — detect "cancel order" intent across 6 languages
• —Multilingual chatbots — build restaurant / catering support bots
• —Cross-lingual transfer learning — train on English, test on low-resource languages
• —Machine translation evaluation — parallel question set across languages
• —Speech / TTS data generation — textual base for synthetic audio
• —Low-resource NLP — Pashto and Sindhi are under-represented languages

🌐 Sample Questions

📊 Dataset Statistics

⚠️ Limitations

• —Small size — 101 questions per language (best for evaluation, not full training)
• —Single domain — All questions focus on restaurant / catering order cancellation
• —Machine-assisted translations — Some low-resource translations may benefit from native review
• —No labels — This is an unlabeled question corpus; add intent labels as needed

📜 License

Apache 2.0 — free for commercial and research use.

🙏 Citation

If you use this dataset, please cite:

@dataset{polyglot_restaurant_questions_2026,
  title  = {Polyglot Restaurant Questions Dataset},
  author = {Nassim},
  year   = {2026},
  url    = {https://huggingface.co/datasets/nassimjp/Polyglot-Restaurant-Questions-Dataset},
  note   = {101 restaurant cancellation questions in 6 languages}
}

🔗 Links

• —Hugging Face: nassimjp/Polyglot-Restaurant-Questions-Dataset
• —Languages: English · Farsi · Japanese · Pashto · Sindhi · Urdu

⭐ If this dataset helped you, please give it a like on Hugging Face!

---

---