Pawel Swietojanski research description

I am interested in various aspects of speech recognition gathered around deep neural networks. In particular I have worked on cross-lingual low-resource speech recognition systems where we applied various strategies to either transfer knowledge between languages in unsupervised fashion or to learn the new features by means of multilingual finetuning. My most recent interests focus on multi-source/multi-channel speech recognisers where I am trying to make the best use of deep neural nets.