Datasets:

Minutor
/

vaani-or-dataset

language:
  - or
license: cc-by-nc-4.0
task_categories:
  - automatic-speech-recognition
pretty_name: Vaani Odia Dataset
viewer: true
extra_gated_prompt: >-
  To access this subset, please provide a link to your HF profile confirming you
  have requested and received access to the original ARTPARK-IISc/Vaani dataset.
extra_gated_fields:
  HF Profile Link: text
  I accept the original ARTPARK-IISc Vaani terms: checkbox

Vaani [Odia (OR) Dataset] Subset

This is a standalone extraction of the Odia language subset from the Project Vaani dataset.

Gated Access & Licensing

To comply with the original data source policies, this repository requires manual approval. By requesting access, you confirm that you have read and accepted the terms of the original ARTPARK-IISc Vaani dataset.

Original Source: ARTPARK-IISc / Project Vaani
License: CC-BY-NC 4.0 (Non-Commercial)

Dataset Details

Language: Odia (or)
Total Splits: Train, Validation, Test

Split Statistics

Split	Utterances	Approximate Duration
Train	24,811	~41.87 Hours
Validation	3,282	~5.50 Hours
Test	2,959	~5.10 Hours

Cleaning Notes

A script-level audit identified 21 utterances across the train and validation splits whose transcripts contained no Odia script (containing Bengali/Assamese, Devanagari, or mixed scripts instead). These were removed to ensure label–audio alignment.

Split	Utterances removed	Audio removed
Train	10	~1m 36s
Validation	11	~1m 50s
Test	0	—

Citation

If you utilize this dataset in your research, you must cite the original Project Vaani creators.