mohdali1/baltivoice-asr

BaltiVoice ASR Dataset

Dataset Description

BaltiVoice is one of the first publicly available Automatic Speech Recognition (ASR) datasets for the Balti language (ISO 639-3: bft), a critically low-resource Tibetic language spoken primarily in the Gilgit-Baltistan region of Pakistan and parts of India (Ladakh).

This dataset was collected, validated, and processed as part of a portfolio research project aimed at building the first open-source Balti ASR system using OpenAI Whisper fine-tuning.

---

Language

Property	Detail
Language	Balti (بلتی)
ISO Code	bft
Language Family	Sino-Tibetan → Tibeto-Burman → Tibetic
Script	Nastaliq (Arabic-based)
Region	Gilgit-Baltistan, Pakistan; Ladakh, India
Speakers	~300,000–500,000 (estimated)
Resource Level	Critically low-resource

Balti is considered endangered by many linguists. It has very limited digital presence, almost no NLP tooling, and until now, no publicly available ASR dataset.

---

Dataset Statistics

Split	Samples	Estimated Hours
Train	9,051	~15.1 hours
Validation	1,006	~1.7 hours
Total	10,060	~16.8 hours

Audio Properties

Property	Value
Format	WAV (16kHz, mono)
Avg Duration	~6.0 seconds
Min Duration	~1.0 seconds
Max Duration	~15.0 seconds
Sample Rate	16,000 Hz

Text Properties

Property	Value
Avg Words/Sentence	10.12
Avg Characters	48.80
Script	Nastaliq (RTL)

---

Dataset Structure

Each sample contains:

• audio: 16kHz mono WAV audio array
• sentence: Balti transcription in Nastaliq script

{
  "audio": {
    "array": [...],          # numpy array
    "sampling_rate": 16000
  },
  "sentence": "بوا لہ سلام بے اِنپا سلام سہ مہ بیاس"
}

---

Source & Collection

• Base data sourced from Mozilla Common Voice Balti (bft) contribution project
• Audio clips were validated for quality and transcription accuracy
• Processed and structured for HuggingFace-compatible ASR training
• Train/validation split applied with random_state=42 (90/10)

---

Usage

Load the dataset

from datasets import load_dataset, Audio

dataset = load_dataset("mohdali1/baltivoice-asr")
dataset = dataset.cast_column("audio", Audio(sampling_rate=16000))

print(dataset)
# DatasetDict({
#     train: Dataset({features: ['audio', 'sentence'], num_rows: 9051})
#     validation: Dataset({features: ['audio', 'sentence'], num_rows: 1006})
# })

Preview a sample

sample = dataset["train"][0]
print("Transcription:", sample["sentence"])
print("Audio shape:", sample["audio"]["array"].shape)
print("Sample rate:", sample["audio"]["sampling_rate"])

---

Model Trained on This Dataset

A fine-tuned Whisper model trained on BaltiVoice is available at:

👉 mohdali1/whisper-small-balti

from transformers import pipeline

asr = pipeline(
    "automatic-speech-recognition",
    model="mohdali1/whisper-small-balti"
)
result = asr("your_balti_audio.wav")
print(result["text"])

---

Social Impact

Balti is an endangered language with very limited computational resources. This dataset contributes toward:

• Preserving Balti language digitally
• Enabling voice technology for Balti speakers
• Supporting NLP research for low-resource Tibetic languages
• Providing a foundation for future Balti TTS, NER, and MT systems

---

Limitations

• Audio collected from volunteer contributors — some variation in recording quality
• Vocabulary may be limited to common conversational domains
• Model fine-tuned using Urdu tokenizer as a proxy (closest supported Nastaliq script in Whisper)
• WER metrics are relative to Whisper's Urdu tokenization, not a native Balti tokenizer

---

License

Creative Commons Attribution 4.0 (CC-BY-4.0)

---

Citation

If you use this dataset in your research, please cite:

@dataset{baltivoice2025,
  author    = {Mohammad Ali},
  title     = {BaltiVoice: A Low-Resource ASR Dataset for the Balti Language},
  year      = {2025},
  publisher = {HuggingFace},
  url       = {https://huggingface.co/datasets/mohdali1/baltivoice-asr}
}

---

Author

Mohammad Ali
BSc Software Engineering, IUB
HuggingFace · GitHub