Toward HMM based machine translation for ASL

Abstract

HMM-based models are widely used in many fields such as pattern recognition, speech recognition or Part-of-speech tagging. However, A HMM can be considered as a simplest dynamic Bayesian network. This network allows us to design a probabilistic graphical model that can be used in machine translation field especially for sign language machine translation. In this paper, we present a Bayesian Learning based method to train the alignment between a simple GLOSS form and a more complicated GLOSS form using sign language specificities such as space locative and classifier predicates.