Spotify has filed a patent application for an AI model that is capable of recognizing plagiarism in songs and lyrics, states Hightech.
As explained in the documentation, the instrument is able to perceive all elements of compositions – notes, melodies, chords, lyrics and additional notes. This model could allow Spotify to immediately check content on a “plagiarism detector” that engineers have trained on tens of hundreds of songs stored on the company’s server.
The main element of the model is a musical plagiarism detection system that can find borrowings based on melodic similarity, sampling or similarity of rhythms. At the same time, musicians will be able to indicate in advance references to certain tracks so that the system does not consider this technique as plagiarism, notes NIX Solutions.
Spotify states that these methods work significantly better than manually selecting tracks. Moreover, the skills of determining plagiarism require significant experience. In this case, the moderators have to listen to several thousand songs; sometimes, the plagiarism may not be heard the first time.
The solution to this problem is the graphical user interface (GUI). It is able to highlight specific pieces of the piece and list the list of tracks that may have been the original source. The moderator only needs to make the final decision. Spotify has yet to say if it plans to implement the system on its platform.