How Pinterest uses Machine Learning?

“With 100+ billion human-curated ideas, Pinterest is the biggest image-rich data set ever assembled. This lets Pinterest do interesting things like analyze trends, understand the intent, and predict consumer behavior.”

With Machine Learning, computers are getting skilled at finding out patterns and features in images and texts. That is how Big Companies like AMAZON and FLIPKART developed complex recommendation systems and TWITTER and INSTAGRAM are making their users stuck to them.

PINTEREST is no exception. It is a visual search platform where people discover and save ideas. Tens of millions of people interact with Pinterest each day, browsing, searching, and discovering ideas inspired by their tastes. To accomplish this, Pinterest engineers used Computer Vision models that analyze the content of each Pin, filtering abusive and misleading content, optimize ad placements, and ranking nearly 300 billion Pins daily.

Now the users love personalization, 80 percent of users are more likely to purchase if the experience is personalized. Analyzing mountains of data by using AI algorithms, Pinterest tailors search results for each of its hundreds of millions of users. Now machine learning touches virtually every aspect of Pinterest’s business operations, from spam moderation and content discovery to advertising monetization and reducing churn of email newsletter subscribers.

“Pinterest at its core, is a data and AI company”
~ Vanja Josifovski, Chief Technology Officer.

In 2015, Pinterest acquired Kosei, a machine learning company that specialized in the commercial applications of machine learning tech (specifically, content discovery and recommendation algorithms).
Kosei’s product allowed customers to make better product recommendations on their sites, apps, and ads. By analyzing what the user had browsed or purchased previously, Kosei’s machine learning engine could compare that against commerce data sets, and predict what someone was most likely to want to buy next.

Pinterest uses Machine Learning in areas like-

• Classification is used to detect spam content and users by The Black Ops team
• The Discovery team provides recommendations, related content, and predicts the likelihood that a person will Pin content
• Visual Discovery team works with complex deep learning algorithms to do object recognition and recommendations
• The Monetization team does ad performance and relevance prediction
• The Growth team has begun to move into the realm of using intelligence models to determine which emails to send and prevent churn
• The Data team is building out a distributed system for machine learning using Spark, so the learning can be efficient and potentially real-time

One way Pinterest makes recommendations is through a neural network called PinSage developed using TensorFlow and PyTorch deep-learning frameworks on Amazon Web Services (AWS). The deep-learning model places each image, according to the, within one giant “graph” of other images.

Three billion images, or “nodes,” form the graph which allows Pinterest to recommend thematically similar images for users. Pinterest’s deep-learning models also learn from what users capture with their phones’ cameras.

VISUAL SIMILARITIES IDENTIFICATION

Machine learning not only determines the subject of an image but also identifies visual patterns and match them to other photos. It helps users to find content that looks like pictures they’ve already pinned.

CLUSTERING

Metadata, such as the names of pinboards and websites where images have been posted, helps the platform understand what photos represent.

LOCAL TASTE

Pinterest recommendation engine suggests popular content from users’ local regions in their native language.

PERSONAL INTERESTS

While many platforms prioritize content from a user’s friends and contacts, Pinterest pays more attention to an individual’s tastes and habits — what they’ve pinned and when — enabling the site to surface more personalized recommendations

CAPTION ANALYZING

Analyzing what’s in a photo is a big factor in the site’s recommendations, but it doesn’t offer the whole story. Pinterest also looks at captions from previously pinned content and which items get pinned to the same virtual boards.