Lexisnexis
Essay Preview: Lexisnexis
Report this essay
Entertainment 2.0
In order to provide a column that is more relevant to you, the reader, please rate this statement on a scale of 0 to 5 (with 0 representing “complete disagreement” and 5 representing “total agreement”):
“Im delighted by the notion that software may soon understand my tastes in music, books, movies, and TV shows better than my closest friends and family members.”
A group of entrepreneurs are hard at work on a new generation of intelligent recommendation software aimed at achieving that. As you turn on the television, hook up your MP3 player to your car stereo, watch online videos, or listen to digital music at work, their vision is that youll always encounter content catering to your tastes – even if its a TV show youve never heard of or a popular band youve never realized you liked.
The urgency behind making the software work better is linked to the explosion of content available online – some of it well-known stuff like the NBC show “The Office” and some of it obscure like the punk album just released by your neighbors garage band.
“Its hard to even keep track of how many videos, TV shows, or books are out there, especially as it has gotten so easy for people to self-publish their stuff on a site like YouTube,” says Steve Johnson, chief executive of Cambridge-based ChoiceStream Inc. “Personalization is a really powerful way for people to discover new content, without having to go out and search for it. With entertainment, people prefer a more passive experience.”
Media companies increasingly look at providing more solid recommendations as a key to retaining customers and getting them to spend more. ChoiceStream supplies its software to DirecTV Inc. to help its subscribers get personalized suggestions online, and it also supports Blockbuster Inc.s DVD-by-mail service.
In October, the dominant DVD rental service, Netflix Inc., said that it was offering a $1 million prize to any person or team that could improve its recommendation software by at least 10 percent. (Netflixs recommendations are pretty poor today; despite my having rated 192 movies on the site, Netflix unaccountably believes that Im a Keanu Reeves fan, suggesting I rent both “Constantine” and “The Lake House.” As far as Im concerned, Reeves is to acting what the Man from Nantucket was to poetry.)
Jim Bennett, vice president of recommendation systems at Netflix, told me “we hear comments like that all the time. Thats one of the reasons we are always trying to improve our recommendations. Our CEO is quite good at finding examples and asking us: `Why did it recommend a fight video when he has just rated a bunch of comedies?”
The science of intelligent recommendations has roots in Cambridge. At the MIT Media Lab, researchers developed software in the 1990s that analyzed how the musical tastes of large groups of people overlapped: If we both enjoy Morphine, which of your other favorite bands might be interesting to me?
In 1995, that software spawned Firefly Network, one of the first recommendation-oriented sites on the Web. Firefly sold its technology to sites like Barnesandnoble.com and Yahoo before being acquired by Microsoft Corp.
Ten years later, the riddle of how to make the perfect recommendation still hasnt been cracked, though some sites seem as if theyre getting close.
Pandora Internet Radio (www.pandora.com), based in Oakland, Calif., has created a site that asks you to name a favorite musician, and then dynamically creates an Internet radio station, programmed only with music from that artist – or one similar in style.
One of the clever things about Pandora is that you dont have to “seed” the service with lots of information about your tastes.
Users