Every restaurant has its signature dish, whether it’s a killer kale salad or an over-the-top chocolate fudge sundae. Unless you’re tuned into the food scene, though, it’s not always obvious which menu item is making waves on Instagram. Luckily, Yelp feels your pain. Today, it announced Popular Dishes, an artificial intelligence-powered recommendation engine that spotlights winning appetizers, entrees, and desserts.
“We’re surfacing the most talked-about things,” Yelp product manager Matt Geddie told VentureBeat in a phone interview. “Yelp’s got a lot of great content, and as we’ve grown over the years … it has become harder for users to digest all of the information.”
Here’s how it works: Yelp’s machine learning algorithms comb through the reviews and pictures on a restaurant’s Yelp page, identifying dishes that are photographed or mentioned at a high frequency relative to others. Those get special mention in the new Popular Dishes carousel, which shows the title of the dish and a prominent picture of it, along with the number of photos and reviews that reference it.
Yelp already shows menus for some restaurants, courtesy of third-party menu providers and delivery partners like Grubhub, and it’s using some of that data to inform Popular Dishes. As Geddie explained, training algorithms partly on user-submitted data allowed it to roll out the feature to restaurants without digitized menus.
“We started a pilot project with just that data [and were] pleased with the results,” he said. “So we’re expanding it to all restaurants [on] Yelp.”
Of course, “popular” isn’t synonymous with “best,” and Yelp isn’t applying sentiment analysis to suss out superior dishes. But according to Geddie, the tastiest menu items naturally bubble to the top.
“We found that most of the time, the dishes that people order the most are usually the better ones,” he said. “We didn’t want to hide stuff from users … [or] penalize restaurants.”
The AI tech underlying Popular Dishes won’t begin and end with food. Yelp is looking into applying it to other business categories, Geddie says.
Popular Dishes comes hot on the heels of Yelp’s Recommended for You, a personalized collection of recommended hotspots informed by business listings a user has viewed and the recommendations they received the previous week.
Yelp’s other forays into machine learning include deep learning-powered image analysis that identifies the color, texture, and shape of objects in photos and a custom ads platform that recommends photo and review content to use in banner ads based on its “impactfulness” with users.