Synthetic Data Can Improve Machine Translation – New Research from IBM, UCSD

New research from IBM, UC San Diego explores synthetic parallel data as a new means of pre-training machine translation models.