Google Geocoding, Ambiguous Locations, and My Maps Submitted Data

Have you ever done a search on Google Local Search like “pizza near empire state building” where you enter a building or a landmark instead of a zip code, or a street, or city or state name?

Google local map with pinpoints for pizza places near the Empire State Building
Pizza Near Empire State Building

While many businesses, organizations, and points of interest (such as parks and schools) have very specific address information associated with them in Google’s Local Search database, people do sometimes want to use landmarks and other more ambiguous locations in their searches, such as neighborhoods (like “pizza in soho”).

A recent patent application from Google pinpoints some of the difficulties that Google’s Local Search may have with searches such as “restaurants near space needle,” where searchers may not be providing much actual geographic information in their searches. It also describes how the search engine might fill in the information it has about locations in its geographic database with user submitted data.

People who use Google Maps can create their own personalized maps and share them with others, but I hadn’t seen anything from Google that stated that they might use information from those maps in Local Search results until this patent filing.

Google local search result for restaurants near space needle
Restaurants Near Space Needle

In my Space Needle search, Google was confident enough about the Space Needle being located in Seattle, Washington that it returned results directly. The patent application discusses how it might create confidence scores for the results of some searches, to decide what it might show to searchers.

When the search engine is less confident about a location for a region or landmark, it might serve a choice of locations, such as the following for “restaurants near times square.”

Google local search result for restaurants near times square showing links to refine the search.
Restaurants Near Times Square

If I click upon the top result for Times Square in New York City, Google enters the following information in the search box – “restaurants loc: Times Square, New York, New York, New York 10036,” which includes City, State, and Zip Code.

The patent application focuses upon how it might find well structured and more complete location information for searches from its database, and from trying to match geographic locations with limited information. The problem it is intended to address is described this way in the document:

However, problems may arise if an address does not follow these conventions. In some countries, addresses typically include references to multiple features and are not hierarchical. For example, some features may include buildings, stores, or landmarks, and the address describes relative proximity to these features. Performing a table lookup is impractical since there is no defined order to the address, there may be a lack of information concerning typical address components, such as roads or street numbers, and a table cannot recognize a description of a spatial relationship between two or more objects.

Geocoding Multi-Feature Addresses
Invented by Apurv Gupta and Tushar Khot
Assigned to Google
US Patent Application 20090177643
Published July 9, 2009
Filed: January 4, 2008

Abstract

A system and method of parsing natural language descriptions of features to determine an approximate location. An embodiment includes splitting the natural language descriptions into components, geocoding each component, and returning the geocode with the highest confidence level. The geocode references a specific location, and this information may be determined by content from a variety of sources. The system may use an assortment of techniques for determining highest confidence level.

When you perform a search that includes geographic information, Google’s Local Search will try to understand more about the geographic region involved in the search such as the name of a city.

It will look in its database, and try to find a match for a location based upon rows and rows and rows (lots of information) of streets and cities and states associated with standard formatted addresses. It might come back from a search like that with a number of results, like in my example above for Times Square. In my other example for the Space Needle, the search engine likely ranked that landmark very highly with a confidence score, and returned only one result. In the times square example, the results shown were likely all possibly reasonable results, and had decent confidence scores associated with them.

When a result is returned with one specific location, the listings that may be shown use that landmark or location as a center point. At the top of my results for “restaurants near space needle” shown above is the line “narrow by Distance | Cuisine | User Rating,” so a number of options are available, but the results are centered around the space needle.

The patent filing also tells us that if Google doesn’t find much helpful geographic information in its database, taken from the many sources that it uses such as yellow page type telecom information and what it discovers on the Web, it may dip into information submitted from people who may have added to Google through their own personalized maps.

Share

9 thoughts on “Google Geocoding, Ambiguous Locations, and My Maps Submitted Data”

  1. just one question from every one expert out there. what is the authenticity of data provided by Google. The sample size of users submitting the data is also very less so that we can’t use as a perfect source for categorizing places.

  2. Hi cousnseling

    Google uses a mix of data sources, from organizations that provide phone listings such as the yellow pages, to online business directories, to websites that contain business location information (possibly including information taken from the sites of the businesses and organizations themselves), to submissions from the owners of listed sites through a verification process.

    If Google also decides to consider information from people who use Google Maps, it’s only after considering information from a wide variety of the sources I’ve listed above. Organizations who might be listed in Google Maps do have some incentive to submit and verify their information with Google, and to have a working and phone directory listed phone, as well as being mentioned and listed in other places on the Web.

  3. Very interesting, I am glad to see more improvements to local search. I am assuming that a large majority of business related searches will eventually have local results displayed.

  4. Hi Joel,

    It’s hard to tell if search engines will start showing local results with every business related search.

    Some non-business related queries may trigger Google Maps, which indexes some content that isn’t business related, such as schools, parks, nonprofits, and government offices.

    Some business related queries may not be worth showing maps upon, because there may not be much of an intent, either express or implicit, that a local search is intended. There are many online businesses that might be as helpful as places with physical storefronts and offices. My guess is that Google may be careful about showing local search results with every query that might be considered business related, especially if there doesn’t seem to be too much desire on the parts of searchers to click through those results. If thousands and millions of viewers see local search results in response to their queries, and no one clicks upon those, a search engine might stop showing those local results.

Comments are closed.