What’s inside this magical black box? It is a type of Recurrent Neural Network (RNN) called an LSTM. An RNN is a deep learning algorithm that operates on sequences (like sequences of characters). At every step, it takes a representation of the next character (Like the embeddings we talked about before) and operates on the representation with a matrix, like we saw before. The thing is, the RNN has some form of internal memory, so it remembers what it saw previously. It uses that memory to decide how exactly it should operate on the next input. Using that memory, the RNN can “remember” that it is inside of an intended scope and that is how we get properly nested output text.