Repository containing projects related to text generation using Artificial Intelligence, a subfield of Natural Language Processing.
The main purpose here is to explore different architectures and datasets, to see if is relatively easy and possible for a simple neural network to capture the structure of a text (i.e poetry) and create new texts.
The very first project in this area is a LSTM that generates poetry based on Machado de Assis' works. It uses 4 of his brilliant books to capture his essence, characteristics and text structure and generates new poetry based on a seed text (a short sentence that is the beggining of the generated poem).
I created a Jupyter Notebook, which runs on Google Colab, that uses the model to generate poetry. It can be found in this link and is very easy to use, but it's in Portuguese for now. The code used to create the model is avaiable here.