Aniemore/REPV

Russian Emotional Phonetic Voices

Crowdsourced phonetic voice clips, five emotions.

How it was collected

REPV was gathered by crowdsourcing rather than in a studio: around 200 different speakers for the full set and about 50 for REPV-S. Recording conditions therefore vary from contributor to contributor, which makes it harder than RESD and closer to what a microphone in the wild actually receives.

Splits

Split	Rows	Hours	Mean clip
train	1628	1.65	3.6 s
test	407	0.40	3.6 s

Fields

Column	Meaning
path	Original file path
file	Source file name
gender	Speaker gender as reported by the contributor
emotion	Emotion label of the recording
speech	Audio

Gender is close to even in train: 883 f, 745 m.

The label set is not the seven-class one used by RESD and the Aniemore models. REPV has five: anger, enthusiasm, happiness, sadness and tiredness. tiredness appears nowhere else in the library, and neutral, fear and disgust are absent here.

Usage

from datasets import load_dataset

ds = load_dataset("Aniemore/REPV")
print(ds["train"][0]["emotion"])

Limitations

Crowdsourced audio varies in microphone, room and level, and the whole set is 2.1 hours — small enough that a single split can move a score by several points. REPV-S in particular holds 2035 clips in total and is meant for smoke tests rather than for measuring anything.

Citation

@misc{Aniemore,
  author = {Артем Аментес, Илья Лубенец, Никита Давидчук},
  title = {Открытая библиотека искусственного интеллекта для анализа и выявления эмоциональных оттенков речи человека},
  year = {2022},
  publisher = {Hugging Face},
  journal = {Hugging Face Hub},
  howpublished = {\url{https://huggingface.com/aniemore/Aniemore}},
  email = {hello@socialcode.ru}
}

License

MIT.