• by Laurie Sullivan  @lauriesullivan, October 24, 2023

Yelp has expanded its use of neural networks to improve ad matching, search results, photo classification and Yelp Waitlist, among other services. 

This means more relevant search results that map back to search intent, ads that are more relevant to them, photos that better depict what they can expect at the business, and more accurate wait-time estimates when using Yelp Waitlist.

Sam Eaton, CTO at Yelp, calls neural networks a key part of Yelp’s “AI toolkit.” He points to the “fast pace of innovation” and Yelp’s ability to “scale neural network training sets.”

The ability to extract complex features from vast amounts of raw data such as review text and photos allows yelp to better understand the queries and intent, along with the needs of business that advertise and list on the platform.

Amazon Web Services (AWS) describes a neural network as a method in artificial intelligence (AI) that teaches computers to process data in a way inspired by the human brain.

Yelp early on adopted neural networks for ad targeting and image classification. The technology is being used to better match the correct ads to a set of consumers on the context of their search query. With support through the technology, Yelp can scale and train the models with data to drive more accurate matching. It can process years worth of data, rather than month, very quickly.

The new models enable Yelp to clearly understand consumer intent and optimize ad placement for businesses to surface more relevant ads.

As part of the ads package, local businesses display photos. They can select them or have Yelp select for them. A new update leverages neural networks that intelligently identifies the content in the photos and then selects the best photos for the businesses’ ads based on factors that match the search query.

The company said it’s been using neural networks for nearly a decade to help categorize photos for restaurants, and more recently food and nightlife businesses.

The ability for this technology to understand “hierarchical features in images” allow people searching for information on the platform to more easily find photos based on what they are most interested in, such as the most recent menu photo or the different types of cuisines.

Yelp uses neural networks to simplify the search and discovery process for those visiting its site. It helps to parse review text to identify food and beverage diners discussed most, and then identifies photos that reference specific menu items.

The photos are used to populate Yelp’s Popular Dishes and Popular Drinks features, which highlight the most highly recommended foods and drinks at a restaurant, cafe or nightlife business.

The company said it is also using neural networks to make the platform safer to use. Neural network-based systems detect photos that may violate Yelp’s content guidelines, flagging images for human moderators to review before they are published on the platform.