How do social networks and other sites define similar music?

Question

Audio

Interested in how, for example, VK, looks for similar tracks with the selected track. What algorithm is used?

Answer 1

you have tracks, Vasya Pupkin has tracks, the sets partially intersect, by combining we will get recommendations for you. Add to this a huge number of users and get something more or less similar to a normal auto-selection of music. If you take into account who listens to what, you can improve the accuracy. Well, etc. This is how most of the music recommendation services actually work. There are other approaches based on the analysis of compositions, but there are not so many such services. Let's say spotify uses echonest to increase sampling accuracy.

Answer 2

The Levenshtein distance algorithm is used. (if it helps you in any way)

Answer 3

Most likely collaborative filtering is applied. Here are a few articles on Habré:
We write a simple recommendation system using the example of Habr
Collaborative filtering
Item-based collaborative filtering with your own hands
In general, look for information on the Recommendation System.