For certain search queries, it is important that documents with regional information are placed near the top of the results. For example, when generating a response to the search query “sushi delivery”, it is appropriate to prioritize results that refer to the user's location. These types of queries are called location-dependent. Depending on the area, they make up about 15–30% of all searches.
Location dependency is determined by statistical data for queries on Yandex. This includes queries that do not contain geographical names, but that are often combined with geographical indicators. For example, the query “order taxi” does not contain a location name, but users often specify which city they need this service in (“order taxi Moscow”, “order taxi St. Petersburg”, and so on). So the query “order taxi” is location-dependent.
To generate relevant output for location-dependent queries, we must determine the user's location and select documents that are related to this area.
The user's location may be detected automatically or set manually. The location is detected automatically based on the IP address. Yandex partners and clients provide data that matches IP addresses with locations. However, the location might be incorrectly identified in certain cases. For example, this may happen if an internet provider offers services in several different regions but assigns similar IP addresses to everyone, without associating them with clients' locations. In these situations, the location should be set manually by the user.
Location dependency is identified while the search query is being processed. Users can also specify to prioritize local results when generating search output.
A site's regional assignment is based on multiple factors, both direct and indirect. For example, these factors may include the text on the site's pages, the site's IP address, and so on.
This technology is currently used for generating search output in Russia, Ukraine, and Belarusia.
Location-based search is a technology that allows us to place regional documents higher up in the search output for a particular class of queries. However, non-regional documents are not excluded from search results, though they may be moved to lower positions.