New Advances in Cross-Task and Speaker Adaptation for Air Traffic Control Tasks

In this paper we explore several options for cross-task adaptation in speech recognition and compare them to develop the new system from scratch. We compare cross-task MAP and MLLR adaptation, and both of them together, in two speech recognizers for air traffic control tasks, one for spontaneous speech and the other one for a command interface. We show how MLLR can even outperform MAP when a big number of transforms is used, how MLLR followed by MAP is the best option, and we also provide some hints of which are the best options to create the MLLR regression class trees. In all cases, we show the effectiveness of means and variance adaptation. For the command interface, we also include the comparison between MAP and MLLR for speaker adaptation using a variable amount of adaptation data.