Modelling Context in Automatic Speech Recognition door