Yandex Maps learned how to search for places by complex queries

Improved search for places based on machine learning and neural networks appeared in Yandex Maps — now the service analyzes all available information about the organization and finds more points for complex requests.

For example, a user can enter “a restaurant with a beautiful view of the city” into the search, and Yandex Maps will show suitable places, even if their description does not mention the view, but it is mentioned in the reviews or it can be seen on photo.

The neural networks used for search analyze not only the description and keywords from the owners, but also the company's website, pictures, menus, price lists and user reviews. So, using the analysis of menus and images, AI will help you quickly find where the right dish is cooked - just ask: “where they cook borscht” or “cafe with desserts”.

Also, the new search can work with more complex cases when people are looking for places with subjective characteristics. For example, "the best hostels in Moscow for two for a long time." Information about this cannot be seen in the photo and can not always be read in reviews or on the site. But the neural network will understand what the user wants, based on all the information available to it.