In order to help users discover new items, recipes, and ideas that relate to their interests, Pinterest is increasingly turning to deep learning — an AI that is fed vast amounts of data through a neural network that simulates how a human brain works until it can autonomously recognize other data.
The platform’s related pins feature pops up beneath a Pin with similar recommendations. In the past these suggestions were surfaced from boards that the same Pin had been saved to by other users. Whereas this method provided an infinite pool of recommendations to play around with, it sometimes resulted in the context of a Pin being lost.
Now, on top of its existing board co-occurrence system, Pinterest is using your activity session data (such as pins you’ve saved or clicked on) along with your recent interactions to label pins, and exclude low-engagement pins. Pinterest claims the complicated deep learning system, dubbed Pin2Vec, helps generate more relevant results that evolve based on user behavior. For example, comparing the two methods in regards to the related pins for a bottle of wine, Pinterest states the following: “Board co-occurrence only found images of bottled wines, whereas Pin2Vec found recommendations for drinks made with wine. This suggests Pinners actually saved the bottled wine Pin and the wine cocktail Pins in the same time series. Pin2Vec helps bridge the gap.”
The company adds that the new method is already proving a success, with early tests showing an increase in engagement with related pins by 5 percent globally. Overall, related pins now account for nearly half of all engagement on Pinterest, again indicating how important the discovery process is on the platform.
“[Pin2Vec] is a big step in improving the relevancy of related pins,” Pinterest wrote in its blog post. “Looking ahead, we’re already making our models faster and analyzing more signals to better personalize recommendations for Pinners around the world.”
