Stochastic Modeling: From Pattern Classification to Speech Recognition and Translation

This paper gives an overview of the stochastic modeling approach in automatic speech recognition and language translation. Starting from the Bayes decision rule for minimum error rate, we present the stochastic modeling approach to speech recognition and analyze its characteristic properties. We discuss the advantages of stochastic modeling and extend it to the translation of written language.