Music Recommendation System Based on Lyrics Similarity and Cosine Similarity

This study describes a music recommendation system that uses TF-IDF vectorization and cosine similarity to propose songs based on lyrical similarities. By converting music lyrics into numerical vectors and assessing song similarities, the system generates textual-relevant recommendations. The model includes an option to filter by artist, which improves tailored recommendations. We used data preprocessing, stemming, and TF-IDF vectorization, along with cosine similarity, to provide music recommendations. The findings show that text-based similarity is beneficial for producing music recommendations. This method could benefit streaming services by pushing personalization beyond user preferences and collaborative filtering.