Abstract
There are many real-world language modelling domains for which there is not a sufficient quantity of training data to reliably estimate a good model. Obtaining enough training data for such "specific" domains can be both costly and a significant logistical challenge. In some cases there may already exist a large quantity of training data from a related or more general domain. The phrase domain adaptation is used to describe ways to adapt models trained on copious, genera data to fit some specific domain well.

In this talk I will show how to do domain adaptation of hierarchical nonparametric Bayesian language models. Specifically I will present a doubly hierarchical Pitman-Yor process language model and explain how such a model accomplishes domain adaptation. I will show results that suggest that this new domain adapting language model performs well in comparison to prior art.