System and methods for optimizing networks of weighted unweighted directed graphs

摘要: Unweighted finite state automata may be used in speech recognition systems, but considerably reduce the speed and accuracy of system. Unfortunately, developing a suitable training corpus for task is time consuming expensive, if it even possible. Additionally, unlikely that could adequately reflect various probabilities word and/or phoneme combinations. Accordingly, such very-large-vocabulary systems often must an unweighted state. The directed graph optimizing methods determine shortest distances between source end nodes weighted graph. These also reweight based on determined distances, so weights are, example, front weighted. searches through are total paths taken will more efficient. Various arbitrarily weight distance reweighting can used.