Personalized Computer Architecture As Contextual Partitioning For Speech Recognition