In this chapter, we introduced one of the most widely used best practices of deep learning. TL is a very exciting tool that you can use to get deep learning architectures to learn from your small dataset, but make sure you use it in the right way.

Next up, we are going to introduce a widely used deep learning architecture for natural language processing. These recurrent-type architectures have achieved a breakthrough in most NLP domains: machine translation, speech recognition, language modeling, and sentiment analysis.