Understanding Seq2Seq Neural Networks – Part 1: The Seq2Seq Translation Problem

There will be problems where we have sequences of one type of thing that need to be translated into sequences of another type of thing . These are called sequence-to-sequence problems , or Seq2Seq problems . One way to solve Seq2Seq problems is with something called an Encoder–Decoder model . To illustrate how to solve a sequence-to-sequence problem using a Seq2Seq Encoder–Decoder model , let's create one that translates English sentences into Spanish . For example, a person wants to say “Let’s go” to his Spanish friend. So we need to create an Encoder–Decoder model that can do this. Let’s look a bit more closely at the problem we want to solve. Not all sentences in the English language are the same length, and the same is true for Spanish . English sentences Let’s go My name is Jack Spanish sentences Vamos Me llamo Jack Also, the Spanish translation of an English sentence can have a different length than the original . For example, the two-word English sentence “Let’s go” translates t