Adaptation of Machine Translation Models with Back-Translated Data Using Transductive Data Selection Methods

Data selection has proven its merit for improving Neural Machine Translation (NMT), when applied to authentic data. But the benefit of using synthetic data in NMT training, produced by popular back-translation technique, raises question if could also be useful data? In this work we use Infrequent n-gram Recovery (INR) and Feature Decay Algorithms (FDA), two transductive methods obtain subsets sentences from These ensure that selected share n-grams with test set so model can adapted translate it. Performing on back-translated creates new challenges as source-side may contain noise originated used back-translation. Hence, finding present become more difficult. Despite that, our show adapting a is an approach.