Employing topic modeling for statistical machine translation

The mixture modeling approaches have dominated the research of domain adaptation in Statistical Machine Translation (SMT). Such approaches construct a general model and several sub-models in advance and focus on the way of determining the relative importance of all the models. In this paper, we propose a simple yet effective approach for better domain adaptation in phrase-based SMT via topic modeling. Different from existing approaches, our topic modeling approach employs one additional feature function to capture the topic inherent in the source phrase and help the decoder dynamically choose related target phrases according to the specific topic of the source phrase. Evaluation on a conversation corpus shows very encouraging results.