README.md

---
# dataset card
# https://github.com/huggingface/hub-docs/blob/main/datasetcard.md?plain=1

language: 
- en
#language_details:
#- "en-US"

pretty_name: "UCLA Speaker Variability Database - subset for SDialog"
tags:
- audio
- speech
arxiv: 2008.03616
license: "cc-by-nc-4.0"
task_categories:
# see list https://github.com/huggingface/huggingface.js/blob/main/packages/tasks/src/pipelines.ts
- text-to-speech
#- speaker-identification

configs:
- config_name: default
  data_files:
  - split: train
    path: "audio/*.wav"


---

# Speakers voice samples from The UCLA Speaker Variability Database


audio recordings from 202 speakers: 101 men and 101 women, aged 18-35 years old (avg 20.5), speaking American English.
Speakers deliver natural speech, not read, for about 30 seconds (around 10 utterances).


Dataset was downloaded from publicly open and link on UCLA lab page:
https://ucla.app.box.com/s/iu8mmvuh16a19e8phjg0ol8gub0zyzf9 

## LICENSE:

in the dataset, and publications it was written:
> "freely available" with attribution.
> COPYRIGHT NOTICE: Please cite our corresponding papers

## References:

- https://assta.org/proceedings/ICPhS2019/papers/ICPhS_785.pdf
- https://linguistics.ucla.edu/people/keating/Keating-etal_ICPhS2019_talk_web.pptx
- https://arxiv.org/pdf/2008.03616
- Ph.d thesis https://www.seas.ucla.edu/spapl/paper/Amber_Afshan_Dissertation.pdf