Our recommendation engine is responsible for personalizing the service to each uses. Our recommendation engine provides:
Our recommendation engine is powered by a project called Gorse, which utilizes modern algorithms. Gorse has very good performance, both in terms of the quality of the recommendations and its ability to generate a high volume of recommendations per second.
The recommendation collects the following data:
The engine uses something called collaborative filtering. The technique comes with recommendations based on how the user has previously interacted with different books, rather that using data about the books themselves. Based on the feedback we get from users, we are able to find users that have a similar taste in literature (they have given similarly positive feedback to the same books), and books that appeal to the same demographics (books enjoyed by similar users).
On the home screen for a user X, the system generates a list of books that users with a similar taste to X have enjoyed. It then filters out the books that X has already interacted with from that list. The remaining books can be recommended to X, and are then displayed on the home screen.
On the finished book screen, the system generates a list of similar books to the one that was just finished, and then filters out the books that X has already interacted with.
The Beat core API exposes two endpoints that generates recommendations:
/v2/releases/{id}/related/releases - which returns releases similar to {id}
/v2/releases/recommended - which returns recommended releases for the currently active user