Swahili Speech Dataset Development and Improved Pre-Training Method for Spoken Digit Recognition

Speech dataset is an essential component in building commercial speech applications. However, low-resource languages such as Swahili lack a resource that vital for spoken digit recognition. For where resources exist, they are usually insufficient. Thus, pre-training methods have been used with external to improve continuous the best of our knowledge, no study has investigated effect specifically This aimed at addressing these problems. First, we developed Then, cross-lingual and multi-lingual on Finally, proposed effective language-independent method The advantage incorporating target language data during stage leads optimal solution when using less training data. Experiments (being developed), English, Gujarati datasets show achieves better performance compared all baselines listed this study.