How Search Engines Might Divine the Intent behind Regional Queries vs. Global Queries
If you search for “pizza,” or “movie times,” or “division of motor vehicles,” there’s a chance that you might want to find information about where to get pizza near you, or to find what films that local movie theatres or showing, or find out more about driver’s licenses in your area. This is true even if you don’t include a specific location with your search.
The query term you used in your search might be considered to be a “regional sensitive query,” because you want to find information associated with a specific geographical location. That geographic location might be on a country or province level, within a specific region, at a state level, or even in a more narrow area such as within a specific city.
How would a search engine decide whether a specific search term might be “regionally sensitive,” or be a “global query” and have no specific local intent behind it?
A recent patent application from Yahoo explores a number of ways that look at user data related to searches to attempt to identify whether a query is regionally sensitive or doesn’t have some kind of location-based intent behind it.
The patent application is:
Identifying Regional Sensitive Queries in Web Search
Invented by Ya Zhang, Srinivas Vadrevu, Belle Tseng, Gordon Guo-Zheng Sun, and Xin Li
Assigned to Yahoo!
US Patent Application 20090307198
Published December 10, 2009
Filed: June 10, 2008
A system for determining the intent of query that includes a search engine that receives a first search query, a query/click log module configured to store log data associated with the first search query; and a computational module that generates metric values associated with the first search query based on the log data and that determines that the first search query is one of a regional specific query or a global query based on the metric values, where the metric values reflect a likelihood of local intent of the first search query, and where the search engine provides search results selected in part based on the metric values.
While the patent filing goes into a fair amount of detail on what it might look for to decide whether the intent behind a query involves a location or not, it can be drummed down to data taken from the search engines query log files, such as:
1. How often a query term like “train tickets” is written to include a geographic location, such as “new york train tickets.”
2. How often a searcher might click upon a check box requesting the search be confined to a specific geographic region, providing a choice between “search the Web,” or limiting the search to something such as “search within Ireland.”
3. How often searchers click upon a geographic-related search result versus a global search result in response to a query.
Google will often show Google Maps results in response to some queries and not others as well, and will sometimes show “customized location” results based upon our previous search history, or aggregated user data from other searchers. It’s possible that Google might be looking at some similar kind of user data.
Added: December 15th – Yahoo announced on their Yahoo Search blog today that they will start showing Yahoo local business results for some queries even when you don’t include a location in a query, in their post: Get More Personally Relevant Results When You Search for Local Businesses. They tell us that we might see local listings for searches such as “yoga” or “auto repair” as well. Are they using the process described in this patent filing to determine whether a query has a location intent? It’s a possibility.