A9 Search Looks to Patent Its Shtick

The A9 search service tries to boost search relevance by taking into account what the searcher did before. Now, it hopes to get a lock on the method.

On Tuesday, the United States Patent & Trademark Office published the Amazon.com subsidiary’s patent application No. 2005003380. “Server architecture and methods for persistently storing and serving event data,” filed in July 2003, describes A9’s method of personalizing search results by including an individual’s past searches and other behavior. The same application is on file in the European Union.

The patent application discloses what is likely already Amazon.com’s de facto search architecture, both for A9 and the e-commerce site.

The online retailer launched a beta version of the A9 portal in April 2004.

Its three-column search results return algorithmic results from Google and from Amazon.com’s own “search inside the book.” The third column, which can be hidden, displays the user’s search history — but only if the user logs in.

In May of that year, Udi Manber, the former chief algorithms officer at Amazon.com who now leads the A9 subsidiary, promised a trade show audience that the company would continue to develop new ways to increase the relevancy of search results. Manber is listed as an inventor of the technology, along with Taylor Van Vleet, Yu-Shan Fung and Ruben Ortega.

The very detailed patent application includes a search application and a server that records which search results were clicked on by the user. It also can tell which results were displayed and which were clicked on during previous searches. The system could flag those URLs the searcher had previously selected.

Searchers could opt to restrict a search to items previously viewed, items not previously viewed or items viewed within a particular time period.

Similarly, the system could highlight which search results were not displayed the last time a searcher used the same query. The system could do the same with things someone clicked on from the product catalog or other Amazon.com services.

For example, search results within an online auction site could indicate which of the located auctions the user had already viewed, when each auction was viewed, and possibly whether the user has submitted a bid on each such auction. A similar system could be used to show which blog or bulletin board postings had already been read.

In the application, Amazon claims that the server architecture described is much more sophisticated than those used by other e-commerce sites. While others track clicks, it states, “these types of records typically lack the level of detail and structure desired for flexibly building new types of real-time personalization applications.”

“They’re talking about a sophisticated event architecture for personalization, which certainly shows how far ahead of the game they are already,” said Guy Creese, a principal in Ballardvale Research, a firm that consults on best online practices for businesses. He noted that the patent describes not only click analysis but also mouseover events . He also said researchers at MIT found that mouse movements are a valuable adjunct to click analysis in predicting Web users’ interest in content.

“An interface might allow searchers to manage and save their searches, the application reads. As mentioned above, users may also be permitted to ‘delete’ specific events from their respective event histories,” it says. The patent application makes clear that these results would not actually be deleted from the servers, but merely eliminated from the search results.

Creese said that while most e-commerce sites don’t have such complicated tracking and personalization schemes, there’s a trend for the biggest, richest competitors to gather highly detailed information about site visitors. ” I can see where [such a patent] could lead to some problems for them,” Creese said.

Slashdotters focused on one item in the patent describing how a searcher can choose not to have results from the search history displayed in the results. Even when they select this option, however, A9 will continue to store their histories, a fact that might not be clear to users.

While Amazon.com doesn’t follow the best practice of asking users to agree to its privacy policy before they register, its tracking practices are clearly disclosed within the site’s privacy policy, available via a link on the main page. A9 will read Amazon.com cookies and make use of that information, it says, and it also may combine it with information from other sources. Those who don’t want to be tracked can use generic.a9.com, an alternative search service that doesn’t track search histories or behavior.

“You can’t fault them on failure to disclose,” said Lee Tien, a staff attorney for the Electronic Frontier Foundation (EFF). “Their honesty is so refreshing that you can see you have no privacy.” The EFF advocates for consumer privacy and the security of online information.

Still, Tien said, the fact that Amazon.com is a seller of books and other information goods and, yet, collects so much user information, is troubling. In his view, the optional nature of A9 search doesn’t ameliorate the harm.

“Part of the problem is that people aren’t all that careful,” he said. “We have a gazillion consumer protection laws that exist precisely because, in the real world, people make a lot of mistakes about this stuff.”

The patent application, which could be granted either soon or within a year or two, only seeks to put a lock on the system Amazon.com likely already uses. If granted, it could slow down similar personalization schemes by search rivals.

An Amazon.com spokeswoman said she couldn’t comment on pending patents.

News Around the Web