A Google researcher says the key to understanding Google's strategy is to look beyond the consumer applications of its patented technology.
The USPTO (U.S. Patent and Trademark Office) published twelve Google patents last week, revealing more details about the company's eforts to employ geocoding technology in its search applications.
The patents (summarized by SearchEngineWatch here) cover topics such as reducing ambiguity in geographic location, weighting the value of the relationship between business name and location, and determining the value of local search results.
Several patents also cover the process of matching ads to page content, automating the advertising approval process and providing real-time transportation data for travelers.
The patents were submitted in the latter months of 2004, but were only published last week by the USPTO. The patents could ostensibly cover recent upgrades to Google Maps, Google Earth and Google's AdWords program, to name a few.
A Google representative declined to say which products the patents covered. "We file patent applications on a variety of ideas that our employees develop," the representative said. "Some of those ideas mature into products and services and some don't."
Google observer, consultant and book author Stephen E. Arnold of Arnold IT said Google typically announces products or makes public upgrades around the time that patents are about to be published.
"They're not new patents," Arnold said. "The patent process is a two- to three-year slog. What Google does is, when the patent process moves from pending to being approved and published, that's when they push the button on doing some serious things. The patent about associating features, that's one more indication that Google is pushing into informed neural networks."
While the technology covered in Google's recent patents applies to ads and consumer local search products, Google will likely apply the technology to other applications, Arnold said. While a patent for local relevance may seem narrowly applicable to maps or local search, the technology could help identify documents that are not explicitly geocoded.
Google Trends, which debuted earlier this year and can segment queries by geographic location, is an explicit example of this capability. But Arnold says there may be more powerful and implicit uses of the technology that the public is unaware of.
"Say somebody writes a document about blowing up subways, and all we know is it came through an e-mail, one of those crappy free e-mail programs," Arnold said. "And they may make a reference to a subway stop near a courthouse, and they make reference to other trains that share the track. There are no explicit geocoding clues there, you see. There are only notions that are shadows of a geocode. So these algorithms that Google is working on can examine and provide a list of cities that have a subway station with multiple trains near a courthouse. You don't need to name the city of New York or the train or the fact that since 9/11 multiple trains share that courthouse station in lower Manhattan. But these algorithms can recurse and iterate the info in that relaxed neural net to offer a list of candidate cities."
Arnold said he could not comment explicitly on how Google may or may not be working with the federal government.
Google is currently hiring federal and Department of Defense sales managers. According to the job descriptions, those salespeople are responsible for generating and closing sales of the Google Search Appliance and Google Earth among U.S. Department of Defense government agencies.
Baseline magazine recently published a cover story that covers, in part, how Google could work with the military.