Project Question of Interest

Given a fixed genre, can the following 7 values be predicted by the song's lyrics and release year?:

• Acousticness, Danceability, Energy, Instrumentalness, Liveness, Valence, Speechiness

The variables are defined as follows:

• Acousticness:
• Danceability:
• Energy:
• Instrumentalness:
• Liveness:
• Valence:
• Speechiness:

Why Care about this Project?

Results

Directory/Links

This GitHub is part of an almost daily series of posts that I have been making on my LinkedIn. Below you will find links to each of the posts grouped by general topic that was discussed. The groups are largely chronological.

Data Gathering/Scraping

Day 1: Kickoff || Day 2: Spotify API || Day 3: Genius API

Data Cleaning/Validation

Day 4: Initial Validation || Day 5: Be Critical and Patient

Data Exploration/Analysis

Day 6: Visualizations of Distributions || Day 7: Finishing Exploration: Correlation Heatmap

Simple Multi-Output Modelling

Day 8: Be Dumb, Create a Baseline!

Vectorization of Lyrics

Day 9: Turning Words Into Numbers: TFIDF || Day 11: Turning Words Into Numbers: Word2Vec || Day 12: Turning Words Into Numbers: Document Embeddings

Motivation/Rationale

Day 10: Why Do This Project

Neural Networks and Transformers

Day 13: Reading NN Documentation || Day 16: Forward Leap - Transformer Embeddings

Don't Stress Out, Life Gets in the Way, Go With the Flow

Day 14: Go With The Flow || Day 15: Go With the Flow (II)

Clustering

Day 17: K-Means Clustering || Day 18: Expectation Maximization Clustering || Day 19: Density-Based Clustering, DBSCAN