Description

In this repository, you will find large-scale pseudo-labeled data, including Central Kurdish audio translated into English. This dataset contains 1.7 million samples, equivalent to 3,000 hours of Kurdish audio, extracted from audiobooks and translated into English using a pipeline that combines a speech recognition system with a machine translation system. The samples have passed several filters, as described in the related paper.

This dataset was developed as part of the COMMUTE project.

Note: Due to the large size of the dataset, publication may take some time. The download link will be shared on this page once it becomes available.