This repo contains all of the code for our final project for COMP 550 - Natural Language Processing. For this project we developed a set of novel features based on rhyme-scheme and rhythmic qualities. We augment several common language models with these features to assess their usefulness in the task of musical genre classification by song lyric. We also built a custom dataset by scraping LyricGenius for the top songs across a small selection of genres.
We also include our final project report. This contains all of the info on project design as well as a discussion of the results.
Navin Kumar, Sung Jun Lee, Namdar Kabolinejad, 2021