Efficiently Reusing Old Models Across Languages via Transfer Learning

Tom Kocmi, Ondřej Bojar

Recent progress in neural machine translation is directed towards larger neural networks trained on an increasing amount of hardware resources. As a result, NMT models are costly t…