Paper: Log-linear weight optimisation via Bayesian Adaptation in Statistical Machine Translation

ACL ID C10-2124
Title Log-linear weight optimisation via Bayesian Adaptation in Statistical Machine Translation
Venue International Conference on Computational Linguistics
Session Poster Session
Year 2010
Authors

We present an adaptation technique for statistical machine translation, which ap- plies the well-known Bayesian learning paradigm for adapting the model param- eters. Since state-of-the-art statistical ma- chine translation systems model the trans- lation process as a log-linear combination of simpler models, we present the formal derivation of how to apply such paradigm to the weights of the log-linear combina- tion. We show empirical results in which a small amount of adaptation data is able to improve both the non-adapted system and a system which optimises the above- mentioned weights on the adaptation set only, while gaining both in reliability and speed.