Simple Fusion: Return of the Language Model
This addresses the challenge of efficiently using monolingual data for NMT, offering a simpler alternative to existing methods like backtranslation, though it appears incremental as it builds on prior integration techniques.
The authors tackled the problem of leveraging monolingual data in Neural Machine Translation by proposing a simple method that combines a fixed language model with a translation model trained from scratch, achieving gains of +0.24 to +2.36 BLEU on four test sets.
Neural Machine Translation (NMT) typically leverages monolingual data in training through backtranslation. We investigate an alternative simple method to use monolingual data for NMT training: We combine the scores of a pre-trained and fixed language model (LM) with the scores of a translation model (TM) while the TM is trained from scratch. To achieve that, we train the translation model to predict the residual probability of the training data added to the prediction of the LM. This enables the TM to focus its capacity on modeling the source sentence since it can rely on the LM for fluency. We show that our method outperforms previous approaches to integrate LMs into NMT while the architecture is simpler as it does not require gating networks to balance TM and LM. We observe gains of between +0.24 and +2.36 BLEU on all four test sets (English-Turkish, Turkish-English, Estonian-English, Xhosa-English) on top of ensembles without LM. We compare our method with alternative ways to utilize monolingual data such as backtranslation, shallow fusion, and cold fusion.