The music streaming platform from the Cupertino Corporation is constantly striving to innovate in its work. It is this policy that makes this streaming service the second in the world. Today, Apple Music holds about 24% of the global music market.
In early 2022, Apple acquired AI Music, a British company that uses artificial intelligence to create soundtracks. The trend of applying artificial intelligence to music streaming services has become massive in recent years. Swedish music streaming giant Spotify has long announced the use of artificial intelligence in its activities.
When we listen to a song, our mind easily understands what kind of sentimental atmosphere it awakens in every person. Of course, when we listen to a song, the sensitivity of different listeners is never the same, and so are different tastes.
But what if we reverse this view? What happens when artificial intelligence does it? AI could definitely help us catalog the vast amount of music a user can find on Spotify, Deezer, Tidal, Apple Music, YouTube Music, and Amazon Music Unlimited. Each of these platforms boasts approximately 30 million songs. It’s not easy to sort them or organize playlists based on mood and related activities: those for the gym, those when we’re sad, or those that should load us.
A research group on a streaming platform with a huge following has begun using artificial intelligence to associate songs with a sentimental world of references.
Automatic detection of musical mood has long been an active area of research for at least twenty years, the study says. The main challenge is to automatically determine the dominant emotion when listening to a song, taking advantage of deep learning and machine learning and including some commonly neglected lyrics in the analysis.
In this work, the Cupertino specialists focused on the task of multimodal mood detection based on audio tracks and song lyrics.
To do this, the researchers built a model based on a set of features. They trained artificial intelligence using audio tracks without lyrics, linguistically reconstructed these songs using a specialized platform, and used a well-known dataset called the Million Song Dataset. First, an attempt was made to collect a large number of descriptive labels. They then attempted to enrich these labels by entering 14,000 English words that attempted to rate them based on the intensity of the emotion being described. These songs were then archived into the music streaming service.
Apple always conducts deep and fundamental research before making a decision to acquire a particular startup and introduce innovations into the operation of a music streaming service. As a rule, the results are always very effective.