The goal for Solemn Simulacrum was to take in a person's facebook data, more specifically their messenger conversation history, and create a program that can generate a response to a prompt similar to what that user would would send. We use Keras’ word processing class to convert the message history into a vocabulary of vectors. When processing a sentence, we convert each word into a vector dimension and concatenate them together to make a sentence vector. This is what is used to evaluate a sentence as the simulated user or a random person. We managed to train a GAN, but we have not been able to generate intelligible sentences.
This project was made to satisfy, in part, the requirements of "COMP 432: Machine Learning" taught by Professor A. Delong.
pip install -U python-dotenv
pip install -U numpy
pip install -U nltk
pip install -U tensorflow
pip install -U keras
Run main.ipynb
, follow the instructions outlining the code cells
Massimo Triassi Evan Dimopoulos
// TODO: Update...
Count Vectorizer to get a word bag count of words used. Naive Bayes to classify text strings. Model Persistence to save the classifier's progress