Common Voice Scripted Speech 26.0 - Caribbean Spanish

Caribbean Spanish Mozilla Common Voice

This dataset was created by filtering the train, dev, and test files from the v26 Spanish Mozilla Common Voice dataset with the following conditions:

• the value of the up_votes field is a number greater than 0.

• the value of the down_votes field is 0.

• the value of the accents column includes "'Caribe: Cuba, Venezuela, Puerto Rico, República Dominicana, Panamá, Colombia caribeña, México caribeño, Costa del golfo de México".

The resulting dataset includes 7,152 clips in the training set, 259 clips in the dev set, and 277 clips in the test set, totaling approximately 12.5 hours of audio.

Data format

The dataset follows the Mozilla Common Voice format: The clips directory contains all of the .mp3 files, and there is a separate tsv file for each data partition, containing the following fields:

• client_id

• path

• sentence_id

• sentence

• sentence_domain

• up_votes

• down_votes

• age

• gender

• accents

• variant

• localesegment

Age Distribution

Gender Distribution