Dynamic Data Selection for Neural Machine Translation
This work addresses efficiency and performance in machine translation, but it is incremental as it adapts an existing technique to a newer model type.
The paper tackles the problem of improving neural machine translation (NMT) by adapting data selection techniques, showing that existing methods yield lower gains for NMT than for phrase-based systems, and introduces dynamic data selection with gradual fine-tuning, achieving improvements of up to +2.6 BLEU over prior selection and +3.1 BLEU over a baseline.
Intelligent selection of training data has proven a successful technique to simultaneously increase training efficiency and translation performance for phrase-based machine translation (PBMT). With the recent increase in popularity of neural machine translation (NMT), we explore in this paper to what extent and how NMT can also benefit from data selection. While state-of-the-art data selection (Axelrod et al., 2011) consistently performs well for PBMT, we show that gains are substantially lower for NMT. Next, we introduce dynamic data selection for NMT, a method in which we vary the selected subset of training data between different training epochs. Our experiments show that the best results are achieved when applying a technique we call gradual fine-tuning, with improvements up to +2.6 BLEU over the original data selection approach and up to +3.1 BLEU over a general baseline.