This project utilizes a character-level Recurrent Neural Network (RNN) trained on tweets posted by @theRealDonaldTrump from 2009 to the present. The training was performed using Lua and a char-rnn model.

To gather the necessary tweet data, we utilized the comprehensive collection available at Trump Archive. The tweet texts were aggregated into a single input file named trainData/input.txt. This file serves as the training data for the character-level RNN.

The training process involved feeding the input.txt file to the character-level RNN, which can be found here. The model was trained for approximately 4 hours. During training, the model was primed with specific texts such as "hello" and "I like" to generate responses. The generated responses, limited to 100 characters each, were aggregated into a file named all.txt within the output directory.

The Facebook Messenger chatbot, accessible here, utilizes the all.txt file as a response database. Upon receiving user input, the chatbot retrieves appropriate responses from the database and delivers them to the user.

Feel free to interact with the chatbot by typing any message. Prepare to be entertained by the amusing and engaging responses generated by the model.

We hope you enjoy interacting with the Trump Facebook Messenger Bot!