Deep Learning with R Cookbook

Neural machine translation

Neural machine translation gained popularity when tech giants such as Google came up with this service. However, this concept has been around for years and is considered one of the most challenging tasks that deals with very sophisticated linguistic models. In this recipe, we will implement an end-to-end encoder-decoder long short-term memory (LSTM) model for translating German phrases into English. This encoder-decoder LSTM architecture is the state of the art approach of addressing sequence-to-sequence (Seq2Seq) problems such as language translation, word prediction, and so on, and is widely used in various industrial translation applications.

Sequence prediction is often framed as an architecture that involves ...