The neural network has finished the “a Song of Ice and fire” by George Martin for
Programmer Zach Tutt (Zack Thoutt) created a neural network that predicts the events of the sixth book of George R. R. Martin’s “Winds of winter”.
The program self-written five chapters — they are published on GitHub. Read more about this tells Motherboard.
The book “Winds of winter” should be the sixth and penultimate part of the Saga “a Song of Ice and fire”. Fans are waiting for the publication of the novel since 2012, however, Martin recently in his blog said that he hopes to release the book only in 2018.
The new part should contain the Chapter telling about the events from the point of view of Sansa stark, Arya stark, Arianna the Sant’agata, Aeron Greyjoy, Theon Greyjoy, Victarion Greyjoy, Tyrion Lannister and Selmy Barristan.
In order to create an alternative continuation of the series, Zach Tutt used LSTM-neural network. It is a subspecies of recurrent neural networks, which are ideal for working with text.
Due to the presence of feedback and ability to learn long-term dependencies LSTM-neural network can work with in the context of long sentences. In addition, it stores information for a long period of time — this helps her to create a coherent narrative and not repeated.