Indonesian-Spanish Learner Corpus

Description

This dataset contains text-to-speech recordings from 14 English language learners with various social backgrounds and dialects among Indonesian speakers.

Specifics

Licensing

Creative Commons Attribution Non Commercial Share Alike 4.0 International (CC-BY-NC-SA-4.0)

https://spdx.org/licenses/CC-BY-NC-SA-4.0.html

Considerations

Restrictions/Special Constraints

The dataset is intended exclusively for research, machine learning development, and internal use. Any commercial exploitation, redistribution for commercial purposes, or use in commercial products and services is prohibited. Users must ensure that the dataset is used only within the permitted scope and in accordance with applicable ethical and usage guidelines.

Forbidden Usage

The dataset is strictly prohibited for commercial purposes. Any use of the dataset for voice cloning, speaker imitation, or the development of systems designed to replicate the identity, characteristics, or behaviour of specific speakers is not allowed. Additionally, the dataset must not be used to train chatbots, large language models, or other AI systems that attempt to simulate or impersonate the original speakers.

Language:

Spanish

Speakers:

There are 14 contributors in total, consisting 13 beginners and 1 intermediate level, from different ages, social backgrounds, and professions. Most of them are native bilingual Indonesian and regional languages (Sundanese and Javanese) of Indonesia. Inside the dataset, the initial folder name indicates the L1 of the speakers, for example Sunda_beginner and Jawa_Beginner.

Sources:

The source text consists of 1,001 sentences taken from this multilingual readability corpus. The sentences are from OpenSubtitles, and are between 1 and 10 words long.

Size:

10 hours with 14 files

Attribution:

Public open access is permitted with proper attribution and citation of the dataset source. Recommended citation:

Abdul Cholik Al Aziz, Diny Opticawati, Masyhuri Farhan, Feby Nur Dianingtyas, Jian Alina, Muqsit Harjuno, Nur Khafidzah, Puput Rizky Afriani, Rahmah Nabilatun N., Sinta Dwi Kurnia, Surti Syafiurahmi, and Yunus Aji Lumintang. (2026). Indonesian-Spanish Learner Corpus [Data set]. Mozilla Data Collective. URL [dataset link].

Structure:

The dataset contains a TSV file, metadata.tsv, with the following columns:

audio_id: A unique identifier for each audio recording, formatted as speaker_id-recording_id.
speaker_id: A unique identifier assigned to each speaker.
audio_filename: The filename of the corresponding audio recording.
sentence: The text corresponding to the audio recording.
num_attempts: The number of recording attempts made by the speaker. Speakers were asked to read each sentence as fluently as possible and were encouraged to retake the recording if they encountered difficulties during the reading.