Skip to content

Dataset · google

Multilingual speech benchmark

google/xtreme_s

Speech benchmark in 102 languages for ASR, translation, classification and retrieval.

What's inside

XTREME-S covers four speech task families: speech recognition, classification, speech-to-text translation, and retrieval. The benchmark covers 102 languages from over 10 language families, 3 domains, and 4 task families. It is designed to evaluate multilingual speech representations and research in universal speech models.

What you can do with it

  • Evaluating multilingual ASR models
  • Testing speech-to-text translation
  • Benchmarking speech classification and retrieval
Tasks
Automatic speech recognition
Languages
Afrikaans, Amharic, Arabic, Assamese, Asturian, Azerbaijani, Belarusian, Bangla and 4 more
Size
10K–100K rows
Downloads this month
20.5K
Likes
73

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("google/xtreme_s")