A patent granted to Microsoft this week hints at the invention of new technology for personalizing searches for online media, particularly music titles.
The patent, granted July 5, is for a method that can train people to analyze media and apply "fundamental properties" that can be matched to computerized analysis such as digital signal processing, or DSP.
"The present invention provides a system and methods for training a trainee to analyze media, such as music, in order to recognize and assess the fundamental properties of any piece of media, such as a song or a segment of a song," according to the patent documentation, originally filed August 21, 2001, by inventors Geoffrey R. Stanfield and Eric Bassman in Seattle. "The process of the present invention includes an initial tutorial and a double grooving process."
Patrick Mahoney, senior analyst with The Yankee Group, says that it makes sense for Microsoft to be exploring new ways to search for digital media content, as the technology has not evolved significantly since Amazon.com began recommending other products to users based on previous purchases.
"This is obviously the next step," Mahoney says. "If you look at the digital content market in general, categorization and search capabilities are in the formative stages at this point. To make more effective digital content services, search technologies need to be much deeper. That's obviously what Microsoft is toying with right now--how do you make those more effective to more consumers?"
The patent states that this so-called double grooving phase would use the skills of experts who have defined a set of classification terms for fundamental properties of music to train others to recognize those properties, and match results of the trainees' assessments with those of experts within "a high degree of tolerance."
"When a high enough degree of cross-listening consensus is reached, the new listener becomes a groover and can classify new songs or segments of songs," according to the patent.
Match My Music
The newly patented training system could be used to train developers to match certain accepted song characteristics--such as volume, tempo, and instrumental use--to DSP technology to design new software, says Jeff Norman, a partner in the intellectual property practice of Chicago-based firm Kirkland & Ellis LLP.
"The person can be trained to converge [his or her] categorization with a DSP analysis so the computer can score you right or wrong," Norman says.
Such technology might be used to personalize an e-commerce user's search for online music on MSN.com, or may be integrated into a highly intelligent search engine that can identify similar music according to how a song actually sounds, Norman says.
"If a user goes to MSN Music and wants to find music that is like Led Zeppelin's 'Black Dog,' how do they find similar music?" Norman says. "Do you look at what other people have ordered, or do you look at other Led Zeppelin [titles]? It seems that Microsoft would like to implement a system where it would come up with characteristics of 'Black Dog' to offer a set of [musical] classifications a user would click on [to find related music]."
Indeed, Microsoft stated in the patent filing that "current classification systems" for search and retrieval of music, such as those used by Amazon.com and BarnesandNoble.com, "are inadequate," which also suggests the software vendor is trying to come up with a better search experience for online media consumers.
Microsoft was not available for comment.
Microsoft has been working on better and more effective search engine technology to integrate this into the next version of Windows, code-named Longhorn, and executives have publicly stated their aim to go head-to-head with current search engine industry darling Google.
The vendor also is waging a battle with a host of companies, including Apple Computer, to sell digital music online.