Using an NLP for emotions dataset to train a variety of models on song lyric emotion content prediction. Includes applications for song recommendations based on an input song's emotional content.
Data on emotions NLP dataset and Spotify songs available in data folder. Fine tuned BERT model and Naive Bayes model class for emotion classification available in model folder. Data preprocessing on emotions NLP dataset contained within emotions_dataset_tfidf.py. Files for the recommender classes are in bert_recommender.py and naive_bayes_recommender.py. Training and test of models and running of recommender done in train_and_test_EmotionsNB.py and train_and_test_EmotionBERT.py
- Clone the repository
git clone https://github.com/mscoop16/spotify_project.git
- Setup python virtual environment and install packages
pip install -r requirements.txt
- If attempting to run flask server, switch to frontend branch
- Make binary into executable and run to start flask server
chmod +x bin/487run ./bin/487run
After setting up, you can use this application at https://localhost:8000. Enter a song title and the number of recommendations you would like to recieve in the form and press submit.
Matthew Cooper, Pratik Nadipelli