Skip to content

Dataset · ai4bharat

Indian-accented English speech for ASR

ai4bharat/Svarah

Audio benchmark with transcripts of Indian English speakers for speech recognition evaluation.

What's inside

Created by the AI4Bharat project to address the shortage of Indian accents in popular speech corpora. Contains 9.6 hours of transcribed English audio read by 117 speakers (1k to 10k examples).

What you can do with it

  • Benchmarking speech recognition (ASR) systems for accent robustness
  • Fine-tuning or adapting models like Whisper for Indian English
Size
1K–10K rows
Downloads this month
3.1K
Likes
61
Access
on request on Hugging Face

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("ai4bharat/Svarah")