Science Behind Spotify’s Discover Weekly



None

Around the early 2000s, Songza executed a manual music recommendation system for all its listeners, where a group of music experts and curators would make playlists. But these suggestions were not the main idea, as they were completely dependent on the personal choice of the curators. It was an average practice for listeners, with a fair amount of hits & misses, as it was impossible to create a playlist which catered to the different tastes of various set of people. The data and technology did not exist back then to create a playlist that would be customized to the taste of every individual listener.

Along came Spotify a few years later, offered a highly customized weekly playlist which is known as the Discover Weekly that promptly became one of their major flagship offerings. Each Monday, millions of listeners got a fresh playlist of new song suggestions, customized along with their personal tastes which were completely based on their listening history as well as the songs they are engaged with. Spotify utilizes a combination of various data aggregation & sorting methods to make their distinctive and powerful suggestion model that is powered by Machine Learning. “One of our flagship characteristics is called Discover Weekly. Each Monday, we give you a list of around 50 tracks that you might not hear before.

 

Spotify uses three forms of recommendation models to power discover weekly:

  

1. Natural Language Processing

NLP is the capability of an algorithm to understand text and speech in real-time. Spotify’s NLP continuously trawls the web to find blog posts, articles, or any other text about music, to come up with a profile for every song. With all this tattered data, the NLP algorithm can categorize songs on the basis of the type of language used to depict them and can match them with other songs that are discussed in the same seam. Artists and songs are assigned to categorizing keywords based on the data, and every term has a certain weight assigned to them. Similar to combined filtering, a vector representation of the song is created, and that is utilized to suggest similar songs.

 

2. Collaborative Filtering

Collaborative Filtering is the most popular method used by recommender systems to create automated predictions about the preferences of all end-users, on the basis of the preference of other similar users. On Spotify, the collaborative filtering algorithm also compares multiple end user-created playlists that have all the songs that end-users have listened to. The algorithm then usually combs those playlists to look at other songs that come out in playlists as well as recommends those songs.

This framework carried out by matrix math in Python libraries. The algorithm first makes a matrix of all the active end-users & songs. The Python library then usually runs a series of multifaceted factorization formulae on the matrix. The end result is two different vectors, where X is the user vector which shows the taste of an individual end-user. Vector Y shows the profile of a single song. To find out users with similar taste, collaborative filtering will evaluate a given end-user vector with every single user vector to give a similar end-user vector as the output. The same method is applied to song vectors. Spotify does not only depend on mutual filtering. The second recommendation model used is NLP.

 

3. Convolutional Neural Networks

Convolutional Neural Networks are utilized to work on the recommendation system and to increase accurateness as less-popular songs might be neglected by the other various models. The CNN model ensures that ambiguous as well as new songs are considered. The CNN model is most commonly utilized for facial recognition, and Spotify has configured a similar model for audio files. Every song is converted into a raw audio file as a waveform. These waveforms are processed by the CNN and are assigned significant parameters for instance; loudness, beats per minute, major/minor key and so on. Spotify then tries to match related songs that have similar parameters as the songs their listeners like listening to.

With these significant machine learning models, Spotify is able to tailor a distinctive playlist of music that completely surprises its listeners almost every week with songs they may not be found otherwise. A major problem in several machine learning models is the lack of access to clean, structured data that can be processed. Spotify has been able to dodge that problem due to their access to enormous amounts of data that they collect from their end-users. They are able to shine as a great instance of efficient use of Machine Learning models to give their end-users an unrivaled personalized experience.

 



Latest posts

AI tools and techniques in Semantic SEO

Semantic SEO are the techniques in search engine optimization which enables the search query to provide meaningful results to the user by understanding the intent behind that query

How AI is Demystifying Traditional SEO Ranking Factors in 2020?

AI is demystifying traditional claims and opinions on 200 Google SEO ranking factors in 2020. Modern-day SEO uses AI tools and techniques in automated content creation, crawling, indexing, semantic topic modeling using word embeddings to enhancing the user experience by page-layout algorithms.  

Top 16 strategic steps in eCommerce SEO and keyword research in 2019

Combine AI tools and techniques in eCommerce SEO with automated content creation and keyword research for eCommerce SEO.

Top 10 AI tools and techniques for automated content creation for SEO in 2019

The essential and ultimate guide to the best top 10 AI tools and techniques to create automated content for digital marketing SEO.

Wordpress Alternative in 2019 | Python Web Development vs PHP web development

WordPress SEO plugins are not so good and people are looking for a WordPress alternative in 2019. Like Python Web Development vs PHP Web Development! What's better, flexible and easy. Let's find out.

Django vs Flask and python based web development in 2019

Django, flask, and CherryPy are one of the best Frameworks for Web Development, so let's try to understand and figure out where Django and flask stands in 2019

Top 8 Factors and Top 8 Solutions for Local SEO in Digital Marketing

We provide top 8 factors which impact search engine results for local SEO based and top 8 solutions to rank your website on the first page of the search engine in 2019.


Coursera: Build New Skills Anytime, Anywhere with 100% online courses. Start Now!