In this project Shakespeare plays by LiamLarsen is used to create a deep learning model which will be able to generate text
using some input text.
While doing all of this we will go through:
Preprocessing
text data- Building multilayer
Bidirectional LSTM
model - Saving
wording embeddings
learned by the learning algorithm
This deep learning model is trained using GPU
and to work in the same environment having packages with versions which were used while making this notebook, go to Kaggle where the kernel is saved.
In Kaggle kernel you'll find embedding vectors and metadata
which can be used to display embedding
learned by the model using Tensorflow Projector
.
The data is text which has 4,583,798
words but only 500,000
words are used train the model. There is no validation
or testing
set. This model is trained for 20 epochs
.
Model's performance by the last (20th) epoch
Text generated by the model
Word cloud for training text