Portuguese - Brazil

Portuguese Speech Data - Spontaneous Dialogue - 87h

Banking
Audio demo
Sender
Invitee

Dataset Details

About

Domain
Banking
Use case(s)
call centre, conversational AI
Model Applications
Acoustic Modelling, ASR Testing, Benchmarking
Total recordings
1408
Hours
87
Word error rate (%) Measurement indicating errors in alignment of text representation (actual vs. perfect) of audio, taking into account words omitted, inserted or wrongly replaced.
4.8%
Average amount of recordings per speaker
30.95
License Type Link
Published date
Sep 1, 2021
File size
15.28GB
Packaging description
A zip file containing metadata files in tsv format and a folder with all the audio files

Demographic

Number of speakers
91.000000
Locale The language(s) and country(s) applicable to the speakers in the dataset.
pt-br
Language
Portuguese
Country
Brazil
Female | Male | Unspecified View on chart
46% | 53% | 1%
20-61
Accent(s) View on chart
Baiano (Bahia), Brasiliense (Brasília), Caipira (Estado de São Paulo), Carioca (Estado e Cidade do Rio de Janeiro), Costa Norte (Ceará), Fluminense (Espirito Santo), Gaúcho (Rio Grande do Sul), Mineiro (Minas Gerais), Nordestino (Rio Grande do Norte, Paraiba, Alagoas, Pernambuco, Maranhão, Piauí), Norte (Roraima, Amapá, Pará), Paulistano (Cidade de São Paulo), Recifense (Cidade de Recife), Sertanejo (Tocantins, Goiás, Mato Grosso, Rondônia), Sul (Mato Grosso do Sul, Paraná, Santa Catarina)

Audio Details

Words
500070
Recording environment
noisy, silent
Audio format
WAV
Bits per sample
8
Device type
mobile
Communication band
narrowband
Sample rate
8kHz

Chart details