It is commonly known that Recurrent Neural Networks (RNN) in theory canmaintain context over a longer period of time, but this has shown to bedifficult in practice. This project aims to expand upon the traditionalimplementations of Long Short-Term Memory (LSTM) RNN in order toachieve a better contextual understanding over a longer period of time,compared to existing implementations. In this report we implement severalprototypes combining an LSTM with other types of networks, with variousminor alterations and for different use cases. The results show that increasedperformance can be gained from this approach, especially on classificationtasks, but that the inherent nature of text generation makes it hard tomeasure the performance of different networks against each others.
Project Gutenberg’s Alice’s Adventures in Wonderland, by Lewis Carroll
This eBook is for the use of anyone anywhere at no cost and with almost no restrictions whatsoever. You may copy it, give it away or re-use it under the terms of the Project Gutenberg License included with this eBook or online at www.gutenberg.org
Title: Alice’s Adventures in Wonderland
Author: Lewis Carroll
Posting Date: June 25, 2008 [EBook #11] Release Date: March, 1994 Last Updated: October 6, 2016
Language: English
Character set encoding: UTF-8
bhaveshoswal, author = Bhavesh Vinod Oswal, title = CNN-text-classification-keras, year = 2016, publisher = GitHub, journal = GitHub repository, howpublished = https://github.com/bhaveshoswal/CNN-text-classification-keras
and
Jason Brownlee: https://machinelearningmastery.com/sequence-classification-lstm-recurrent-neural-networks-python-keras/ https://machinelearningmastery.com/text-generation-lstm-recurrent-neural-networks-python-keras/