Goal of this application is to present our work and capabilities of neural networks to learn language concepts and generate new poems based on given word.
Created neural network is composed of two LSTM layers and is learning text by characters. Learning dataset is composed of 597 poems from American poet Emily Dickinson (With small changes was this datased taken from kaggle 597 poems by Emily Dickinson).
Application was created using python. Interactive widgets were created with help of python library ipywidgets and work of turning it into standalone web application was done by Voilà. To host this application we use Heroku server.
- we are cleaning poetry datasets from Charles Bukowski and William Shakespeare -- we would like to train our model on this datasets as well
- we created model that is learning text by words -- we would like to compare this model with model learning by characters