The Isolated Afrikaans Child Speech (ISACS) dataset contains recordings of isolated words spoken by Afrikaans-speaking children between the ages of four and six. It is designed for few-shot learning tasks and includes adult reference examples for comparison.

• The development (dev) and test sets each contain 16 keyword classes, with a balanced number of positive and negative samples for each class.

• For each keyword, the dataset provides 15 adult template recordings from three different speakers (sp_1, sp_2, and sp_3).

• It also includes 15 child template examples per keyword, spoken by children not included in the dev or test sets.

This setup allows for direct comparison between child and adult speech representations in few-shot classification tasks.

The dataset is described in the following paper. Please cite the paper if you use the data:

• R, Smit, R, Louw, and H, Kamper, “Towards few-shot isolated word reading assessment,” accepted to the Workshop on Speech and Language Technology in Education (SLaTE), 2025.[arXiv]

Download

ISACS (36.84 MB): isacs.zip
MD5 checksum: e3e9c5455c60c3ea40a8d14165765be1

License

© 2025 Stellenbosch University
This data is released under a Creative Commons Attribution-ShareAlike license (CC BY-SA 4.0).