Spotify is the most well-liked music-streaming app on the planet. Along with songs and albums, Spotify has a fantastic assortment of podcasts and discuss exhibits. They’ve just lately launched audiobooks of their app. Like some other providing, Spotify needed to make sure that its audiobook suggestions catered to consumer’s preferences. Therefore, they developed a Graph Neural Community-based suggestion algorithm to personalize audiobook suggestions.
This text discusses the challenges Spotify confronted in delivering personalised audiobook suggestions and the exploratory information analyses performed to deal with them. It explores Spotify’s modern answer: a two-tower graph neural community mannequin designed to reinforce audiobook personalization.
Challenges
As audiobooks had been a current addition to Spotify’s content material library, they confronted some challenges —
- There was a knowledge shortage concern because the content material kind was newly launched. There have been fewer consumer interactions for audiobooks in comparison with different content material sorts. Many customers had been unaware of audiobooks on Spotify.