Welcome to our Deep Learning (CSCI 1470) final project! To learn more about this project, see our poster in poster.pdf. Our dataset (a collection of haikus web-scraped from Reddit) can be found in data/haiku.csv. Our three model architectures can be found in code/naive_n_gram.py, code/line_by_line_no_stops.py, and code/line_by_line_with_stops.py. Scripts to train these models can be found in code/scripts. Once models are trained, poems can be generated using the code/generate_poems functions.
To download our models, begin by creating a directory at code/models. Then click here to access the model files. Download the zip file, and extract it into the code/models folder. Please be mindful that the zip file is approximately 440 MB. Now you can use our three models to generate haikus!