This is a music recommender that aims to reflect the user's thoughts and feelings. It's based on the content of songs and the emotions corresponding to songs and places less emphasis on the audio features so that users get more variety.

Overview

Recommender integrates topic modeling (NLP) and classification.

File Directory

Emotion Classification

• emotion_modeling_setup.py: format datasets for modeling
• functions/classification_functions: contains functions used in modeling for classifiers
• gather_lyrics_data.py: obtain song lyrics
• models/modeling_calm.py: classifier designating if a song is "calm"
• models/modeling_energetic.py: classifier designating if a song is "energetic"
• models/modeling_happy.py: classifier designating if a song is "happy"
• models/modeling_sad.py: classifier designating if a song is "sad"
• predict_emotions.py: apply classifiers to new data

Topic Modeling

• functions/spotify_api_functions: contains functions used to obtain data from Spotify's API
• gather_emotion_data.py: obtain & clean dataset with emotion tags, obtain audio data (to use as features in classifiers)
• topic_modeling.py: topic modeling on lyrics

Recommendor

• recommender.py: code for recommendation system