FLEURS
A multilingual speech recognition benchmark dataset used to evaluate model recognition performance across a wide range of languages.
Timeline (last 90 days)
In the same fine-tuning, FLEURS English WER changed only from 11.04% to 10.42%, and FLEURS Arabic from 12.67% to 11.41%, so the original language capabilities degraded hardly at all.
The study validates its conclusions on the FLEURS and CoVoST2 datasets across three language directions.
FLEURS is a multilingual speech dataset released by Google, covering 102 languages.
All Facts (3)
In the same fine-tuning, FLEURS English WER changed only from 11.04% to 10.42%, and FLEURS Arabic from 12.67% to 11.41%, so the original language capabilities degraded hardly at all.
50%UnverifiedThe study validates its conclusions on the FLEURS and CoVoST2 datasets across three language directions.
50%UnverifiedFLEURS is a multilingual speech dataset released by Google, covering 102 languages.
50%