Dataset · laion
Overlapping speech recognition benchmark
laion/small-overlapping-speech-bench
A stress test for ASR models consisting of 100 audio clips where three people speak different languages simultaneously.
What's inside
A small reproducible benchmark from LAION containing 100 audio recordings in German, French, Spanish, Italian, Dutch, and Portuguese. In each clip, three people speak different European languages simultaneously, and the markup includes timestamps, language, and exact transcription for each speaker.
What you can do with it
- Evaluation of speech recognition (ASR) system robustness under the cocktail party effect
- Testing speaker diarization and language identification models with overlapping voices
- Tasks
- Automatic speech recognition
- Languages
- German, French, Spanish, Italian, Dutch, Portuguese
- Size
- up to 1K rows
- Downloads this month
- 102.9K
- Likes
- 2
How to load it
from datasets import load_dataset
ds = load_dataset("laion/small-overlapping-speech-bench")