This is one of the projects from Udacity's Natural Language Processing Nanodegree where I built a deep neural network, using Keras, that functions as part of an end-to-end automatic speech recognition (ASR) pipeline! The algorithm first converts any raw audio to feature representations that are commonly used for ASR. Then it tests several different neural networks on their ability to map these audio features to transcribed text.
Refer report.html for the overall results, VUI Notebook for the main code, sample_models for the model architecture code.
NOTE: If you are prompted to select a kernel when you launch a notebook, choose the Python 3kernel.