Malfaro43/

whisperedAudio-benchmark

This dataset is an ASR benchmark built from Expresso speech and paired EARS regular/whisper sentence recordings. The benchmark compares ASR performance on the same utterances spoken in two conditions, including background-speaker variants:

Task Condition Expresso style

1 normal or regular speech Expresso default, EARS regular

2 whispered speech Expresso whisper, EARS whisper

3 normal or regular speech with two background speakers at 10%… See the full description on the dataset page: https://huggingface.co/datasets/Malfaro43/whisperedAudio-benchmark.