sulabhkatiyar/ne-asr-dataset-grt-aug

NE ASR Augmented Dataset -- Garo (grt)

Augmented automatic speech recognition dataset for Garo (grt), a Tibeto-Burman language spoken in Meghalaya, India.

Source

Augmented from sulabhkatiyar/ne-asr-grt (original transcribed speech data from the ARTPARK-IISc Vaani project).

Language Information

Property	Value
Language	Garo
ISO 639-3	grt
Family	Tibeto-Burman
Region	Meghalaya, India
Tonal	No
Tier	E (47.33h original data)

Dataset Statistics

• Original training samples: 33,480
• Augmented training samples: 234,360 (7x augmentation)
• Train shards: 670
• Estimated original duration: ~47.3 hours
• Estimated augmented duration: ~331.3 hours

Split	Samples
train	234,360
validation	4,253
test	4,101

Transformations Applied

Each original training sample produces 7 samples (1 original + 2 speed + 4 pitch):

• Speed perturbation: 0.9x, 1.1x (2 variants per sample)
• Pitch shift: -2, -1, +1, +2 semitones (4 variants per sample)
• Noise augmentation: Not applied

SpecAugment Parameters (for training, NOT in this dataset)

These parameters are consumed by the training script and are not baked into the audio files:

• mask_time_prob: 0.12
• mask_time_length: 12
• mask_feature_prob: 0.06
• mask_feature_length: 10
• layerdrop: 0.05

Full augmentation config: configs/augmentation_config.yaml

Dataset Format

• Audio: 16kHz mono WAV (stored as Parquet with audio bytes)
• Text: Transcriptions
• Features: audio, text, language, augmentation
• Augmentation labels: original, speed_0.9, speed_1.1, pitch_-2, pitch_-1, pitch_+1, pitch_+2

How to Use

from datasets import load_dataset

# Load the full dataset
ds = load_dataset("sulabhkatiyar/ne-asr-grt-aug")

# Load only the training split
train = load_dataset("sulabhkatiyar/ne-asr-grt-aug", split="train")

# Filter to only original (non-augmented) samples
original_only = train.filter(lambda x: x["augmentation"] == "original")

# Filter to a specific augmentation type
speed_09 = train.filter(lambda x: x["augmentation"] == "speed_0.9")

Original Data

• Source dataset: sulabhkatiyar/ne-asr-grt
• Project: ARTPARK-IISc Vaani
• License: CC-BY-4.0

Citation

If you use this dataset, please cite the Vaani project and acknowledge the augmentation pipeline.