Transfer learning improves quality for low-resource machine translation, but it is unclear what exactly it transfers . Word embeddings play an important role in transfer learning, particularly if they are properly aligned . Transfer learning can eliminate the need for a warm-up phase when training transformer models in high resource language pairs, say the authors . Even randomly generated sequences as parents yield noticeable but smaller gains, the authors say . They conclude that transfer learning can be used to improve low resource language translation efficiency and reduce the need to warm up training models in languages with high resource languages, such as Arabic and Spanish . The authors conclude that transferring only the embeddents but nothing else yields catastrophic results .

Links: PDF - Abstract

Code :

None

Keywords : transfer - learning - resource - authors - translation -

Leave a Reply

Your email address will not be published. Required fields are marked *