Resampling Approach for Instance-based Domain Adaptation from Patent Domain to Newspaper Domain in Statistical Machine Translation

In this paper, we investigate a resampling approach for domain adaptation from a resource-rich domain (patent domain) to a resource-scarce target domain (newspaper domain) in Statistical Machine Translation (SMT). We propose two resampling methods for domain adaptation in SMT: random resampling and resampling for instance weighting. The random resampling randomly adds sentence pairs from the resource-rich parallel corpus to the target-domain parallel corpus. Instance weighting is a method which provides a weight to each sample in the resourcerich domain. The problem of instance weighting in SMT is how to provide a weight to each sentence-pair. We approximate the instance weights by resampling sentence-pairs according to the ratio of sentence-pair probabilities between the two domains. We also explore a method of selecting samples that have instance weights larger than some threshold.