FSA/FST algorithms, differentiable, with PyTorch compatibility. Automatic speech recognition
Speaker diarization, speake segmentation,