It's easy to get lost in the black hole of Pinterest Related Pins: One DIY project begets a hundred DIY projects begets a hundred DIY projects, and so on—until it's midnight and you still need to feed the cat.
But "related" often seems a relative term for Pinterest, which previously used what it calls board co-occurrence to generate suggestions. Basically a pin is saved to many boards, so the related pins are drawn from that subset of boards.
The digital pinboard, however, is stepping further into the 21st century by adopting a deep learning system to make Related Pins "even more relevant," software engineer Kevin Ma promised.
"We developed a scalable system that evolves with our product and people's interests, so we can surface the most relevant recommendations," he wrote in a blog post.
For years, Related Pins have been generated based on other boards to which they were pinned. But while board co-occurrence provides infinite candidates to recommend, it can also miss the mark.
"Board co-occurrence can lose the context of a Pin, so we needed a way to better understand pins and their relative relationships to pinners," Ma said, describing the Pin2Vec approach to embed pins in the context of users' activity.
Related
- Let People Know You Avoided a #PinterestFailLet People Know You Avoided a #PinterestFail
Pins are labeled and processed through the deep learning system, which relies on a set of algorithms to classify and then select more relevant content: An engaged pin of cuddling lions results in five more images of cuddling lions; an engaged pin for a bottle of wine recommends five drinks made with wine.
While Pin2Vec is "a big step in improving the relevancy of Related Pins [and] an important source for generating candidates," Ma said, it does not replace board co-occurrence.
"Looking ahead, we're already making our models faster and analyzing more signals to better personalize recommendations for pinners around the world," he added.