The reader should now have a good understanding of several machine translation approaches and how neural translation machines are different than their traditional counterparts. We should also now have gained an insight into how to build a neural machine translation system from scratch and how to extend that system in interesting ways. With the information and implementation demonstrations provided, the reader is advised to explore other parallel corpus datasets.
In this chapter, we defined embedding layers but didn't load them with pretrained embeddings, such as GloVe, FastText, and so on. The reader is advised to load the embedding layers with pretrained word vector embeddings and see whether this yields better results. In Chapter 4, Style Transfer in Fashion Industry using GANs, we are going to work through a project related to style transfer in the fashion industry...
