Music is always indispensable in people’s daily life, from the long-existing live performance of various types of instruments to the recording industry that has experienced prosperity and decline in the past years, music has been communicated globally through different mediums. The earliest form of vinyl record evolved to tapes and Cds in the 80 ’s, which has evolved to streaming recently. Throughout the evolution, the relationship of music itself with, not only music producer but also the audience, has dramatically changed.

Obvious changes have appeared on the audience. In the past when information was not as easy to be reached as today, people could only discover music through record store, magazine, friends’ recommendations or radio, the number of channel was rather limited. However, in the Internet era, people can not only search and listen to music online by themselves but also reach music via an unprecedented approach that was created by streaming media.

Spotify, a music streaming service developed by Swedish company Spotify Technology, has utilized Machine Learning to analyze its users’ listening behavior and its song database. It first assigns a certain vector to each of its users and songs first, which represents a user’s favorite and a song’s characteristics, then cooperates with collaborative filtering to further analyze the characteristics of the songs in its database and its users’ taste on music. The result of these two processes is implemented in the software to recommend users the songs that they possibly like. Moreover, Spotify also applies convolutional neural networks to analyze raw audio data, which allows them to have a better understanding on the songs’ characteristics and capture the fundamental similarities between them, therefore provides song suggestions based on the similarities of the song that a user has listened. Spotify is just one example of the many music streaming media, it is not hard to believe that other companies also have comparable methods to precisely provide song recommendation via Machine Learning.

On the one hand, audiences benefit greatly from the advantages of technology, they’re able to enjoy their favorite music that really fits their taste without making much effort on searching. On the other hand, people used to spend a lot of time on listening to different music in a record store before finding the pleasant ones, the absorbing process of which is part of the joy in searching music. Yet the process has been indirectly eliminated by the precise recommendation of music streaming service, and people might less appreciate good music simply because they can easily find some. Is technology promoting music communication or diminish the value of music? Who knows.

Source: https://medium.com/s/story/spotifys-discover-weekly-how-machine-learning-finds-your-new-music-19a41ab76efe