An Automatic Speech Recognition Application Framework for Highly Parallel Implementations on the GPU