In this case study we explore two frameworks for speech recognition: CMU Sphinx and Kaldi. Given varying dependencies, we split them into two separate Docker images and handel them separately. Both methods leverage the Common Voice and contain their own README for instructions.

The CMUSphinx case study trains a speech recognition model using GMM/HMM models. The Kaldi case study follows the common voice recipe, scripted in a jupyter notebook for closer inspection.

More information can be found at: Deep Learning for NLP and Speech Recognition by Springer