Skip to content

Dataset · davidai-labs

DAI-ASR-I18N audio benchmark

davidai-labs/DAI-ASR-I18N

An audio corpus of everyday dialogues in 21 languages for speech recognition and speaker diarization.

What's inside

The benchmark from David AI includes mismatched audio recordings of ordinary two-person dialogues across 21 languages. Each speaker is recorded on a separate channel, and the clips contain human-verified verbatim transcripts broken down by speaker.

What you can do with it

  • Evaluating and benchmarking ASR (Automatic Speech Recognition) models
  • Testing speaker diarization algorithms in a multilingual environment
Tasks
Automatic speech recognition
Languages
Arabic, Bangla, German, English, Spanish, French, Hindi, Indonesian and 4 more
Size
1K–10K rows
Downloads this month
1K
Likes
21
Access
on request on Hugging Face

How to load it

from datasets import load_dataset

ds = load_dataset("davidai-labs/DAI-ASR-I18N")