A better algorithm

Music Similarity

How do you guage similarity of distinct music pieces? Sure you can ask an advanced machine learning model but how does the model figure it out?

A considered approach would first catalog musical elements that give each composition its character and style. There are several elements but here we have focused on the following:

  • Texture. Heterophony, Homophony, Monophony, Polyphony.
  • Quality. Major, Minor
  • Tempo. Adagio, Adagio Sostenuto, Alla Marcia, Allegretto, Allegro, Allegro Agitato, Allegro con Spirito, Allegro Giusto, Allegro ma non Troppo, Allegro Moderato, Allegro non Molto, Allermande, Andante, Andante très Expressif, Andantino, Andantino con Moto, Andantino Sognando, Grave, In moto Tranquillo, Sereno, Largo, Lento, Moderato, Moderato Assai, Molto Adagio, Molto Allegro, Poco Moto, Prestissimo, Presto, Presto Agitato, Sostenuto, Tempo Giusto, Tranquillo, Très Modéré, Vivace
  • Time Signature. 12/8, 2/2, 2/4, 3/4, 3/8, 4/4, 6/4, 6/8, 8/8, 9/8

Once each composition has been assigned an appropriate value for each known musical element, they can be compared to determine the number of element values that match across compositions. The higher the number of matches, the more similar these compositions are considered.

Similar composers? No problem.

Composer similarity could be determined by aggregating the number of similar compositions.