This repository is a test-only release of the EmoS speech-dialogue data. It
contains no training audio or training metadata.
archives/test/cn.tar and archives/test/en.tar: bilingual standard
evaluation data with 1,020 scenarios and 4,080 scored model responses.
archives/test50_human/cn.tar and archives/test50_human/en.tar: blinded
human-evaluation packages for the same 1,020 scenarios and 4,080 WAV files.
All TAR archives are uncompressed… See the full description on the dataset page:
https://huggingface.co/datasets/zsy814/EmoS.
