VP Jim Bennett discusses how recommendation systems suggest your next movie and the challenges of building a better one.
By Kate Greene
Earlier this week, Netflix, the online movie rental service, announced it will award $1 million to anyone who can come up with an algorithm that improves the accuracy of its movie recommendation service.
In doing so, the company is putting out a call to researchers who specialize in machine learning--the type of artificial intelligence used to build systems that recommend music, books, and movies. The entrant who can increase the accuracy of the Netflix recommendation system, which is called Cinematch, by 10 percent by 2011 will win the prize.
Recommendation systems such as those used by Netflix, Amazon, and other Web retailers are based on the principle that if two people enjoy the same product, they're likely to have other favorites in common too.
But behind this simple premise is a complex algorithm that incorporates millions of user ratings, tens of thousands of items, and ever-changing relationships between user preferences.
To deal with this complexity, algorithms for recommendation systems are "trained" on huge datasets. One dataset used in Netflix's system contains the star ratings--one to five--that Netflix customers assign to movies. Using this initial information, good algorithms are able to predict future ratings, and therefore can suggest other films that an individual might like.
Because access to such a dataset is critical to improving the quality of its recommendation systems, the company also released 100 million recommendations--stripped of any personal identifying information--according to Jim Bennett, vice president of recommendations systems at Netflix.
We spoke with Bennett this week about how recommendation systems work--and the challenges of building a better one.
Technology Review: Before building a better recommendation system, it would be useful to understand your current approach. How does Cinematch work?
Jim Bennett: First, you collect 100 million user ratings for about 18,000 movies. Take any two movies and find the people who have rated both of them. Then look to see if the people who rate one of the movies highly rate the other one highly, if they liked one and not the other, or if they didn't like either movie. Based on their ratings, Cinematch sees whether there's a correlation between those people. Now, do this for all possible pairs of 65,000 movies.
TR: So Cinematch would recommend movies to me based on the evaluations of people who rated movies to the way I did. Does that method work for all movies at Netflix?
JB: A lot of the really obscure discs, for instance, the "How to Mow a Lawn" DVDs, don't have very many ratings and this method doesn't work as well. For movies with a large number of ratings, you do substantially well. But to make it work, there needs to be a lot of data-tuning because people can sometimes have interesting rating patterns.
TR: Like what?
JB: For example, there are many people who rate a movie with only one star or five stars. And there are some people who just rate everything with three stars. What you're looking for is an interesting spread of opinions because you're trying to capture correlations. That's the core of the engine. Page 2
Monday, October 09, 2006
Windows Vista RC2: Near final OS more refined, but it's not perfect
Microsoft continues to tweak its upcoming operating system
Scot Finnie October 08, 2006 (Computerworld)
Last Friday, Microsoft pushed Windows Vista Release Candidate 2 (RC2), build 5744, out the door. There's no laundry list of new features and functions associated with this build of Vista, but there is something significant about it.
You can sum it up in one word: refinement.
The areas of installation, performance and bugginess associated with Media Center and the new Sleep power management mode have all been improved. Vista continues to be exceptionally stable. There's no question that this new Windows is more reliable than XP.
Setup certainty
I installed RC2 on three test machines, one Windows XP Pro upgrade and two clean installs. One upgrade installation screen notes that your "upgrade could take several hours" to complete. Mine didn't, although it did run a little over an hour and a quarter. The two dual-boot clean installations were swifter than with Release Candidate 1, and showed some minor visual changes.
The results of RC2's setup process were a tad cleaner. My three test machines were manufactured in different years -- 2003, 2005 and 2006. The two newer models are laptops that have proprietary software for controlling hardware. Although all three have hardware Vista was unable to provide drivers for, with just a couple of exceptions, the new OS quickly accepted "legacy" drivers designed for XP.
Vista's driver pack support for recently released hardware continues to be a weak point. I had expected that, with this release, the driver pack would be better than it is. For example, Vista RC2 lacks a driver for the Linksys EG1032 Gigabit PCI NIC. It also wasn't able to locate a driver for SoundMax audio cards on my oldest and newest machines. SoundMax audio is widely distributed and there's just no excuse for this. It was easy for me to find and feed Vista my OEM-provided XP drivers for these devices, however.
Not so easy to get around is the fact that Vista doesn't contain drivers for IBM/Lenovo's UltraNav built-in pointing devices (which have been shipping with ThinkPads for years). Nor was Vista able to run these XP drivers, even when I used some of Vista's compatibility tricks.
continued 2 3
Scot Finnie October 08, 2006 (Computerworld)
Last Friday, Microsoft pushed Windows Vista Release Candidate 2 (RC2), build 5744, out the door. There's no laundry list of new features and functions associated with this build of Vista, but there is something significant about it.
You can sum it up in one word: refinement.
The areas of installation, performance and bugginess associated with Media Center and the new Sleep power management mode have all been improved. Vista continues to be exceptionally stable. There's no question that this new Windows is more reliable than XP.
Setup certainty
I installed RC2 on three test machines, one Windows XP Pro upgrade and two clean installs. One upgrade installation screen notes that your "upgrade could take several hours" to complete. Mine didn't, although it did run a little over an hour and a quarter. The two dual-boot clean installations were swifter than with Release Candidate 1, and showed some minor visual changes.
The results of RC2's setup process were a tad cleaner. My three test machines were manufactured in different years -- 2003, 2005 and 2006. The two newer models are laptops that have proprietary software for controlling hardware. Although all three have hardware Vista was unable to provide drivers for, with just a couple of exceptions, the new OS quickly accepted "legacy" drivers designed for XP.
Vista's driver pack support for recently released hardware continues to be a weak point. I had expected that, with this release, the driver pack would be better than it is. For example, Vista RC2 lacks a driver for the Linksys EG1032 Gigabit PCI NIC. It also wasn't able to locate a driver for SoundMax audio cards on my oldest and newest machines. SoundMax audio is widely distributed and there's just no excuse for this. It was easy for me to find and feed Vista my OEM-provided XP drivers for these devices, however.
Not so easy to get around is the fact that Vista doesn't contain drivers for IBM/Lenovo's UltraNav built-in pointing devices (which have been shipping with ThinkPads for years). Nor was Vista able to run these XP drivers, even when I used some of Vista's compatibility tricks.
continued 2 3
Wednesday, October 04, 2006
Startup Co. Plastic Jungle Offers Gift Card Exchange
A new Web gift card exchange could shake up the gift card space, with cards selling to consumers for less than face value and retailers getting the gift that keeps on giving: previously unavailable data on who is using these cards.
Trying to address the hundreds of millions of dollars lost each year in gift cards that are never redeemed, Plastic Jungle, of Fresno, Calif., has created a Web gift card exchange with a twist: The company is selling valid gift cards for less than the card's face value, which could have a significant impact on retail gift card sales.
Plastic Jungle CEO Tina Henson said the company plans on partnering with major retailers, to help them reduce accounting losses from unredeemed gift cards.
"Home Depot had $43 million gift cards that were unredeemed and more than two years old," said Henson. "It just goes on their books as a liability because they're not able to show it as revenue until it's been redeemed."
The way it works is consumers go to the site and fill out some forms (the current site doesn't ask much but that's going to change by mid-October), and they can then sell whatever gift cards they have for something in the range of 65 to 75 percent of the card's remaining value. The site charges a flat fee of $3.99 for each gift card listed, regardless of value, and all transactions are processed through PayPal.
Consumers can also purchase gift cards and pay about 90 to 95 percent of the card's remaining value, with PlasticJungle.com keeping the difference, Henson said. This would allow consumers to purchase, for example, a $100 Home Depot gift card for $90.
The actual percentages paid and received will vary based on supply and demand, Henson said, adding that some gift cards are simply more popular than others and therefore demand a higher percentage. What are the most popular gift cards? Those from Home Depot, Wal-Mart and Target. The least popular? Barnes & Noble and Borders.
Henson speculated on a few reasons for those popularity differences, but thought a key reason are that those booksellers happen to still use expiration dates (two years for Borders and one year for Barnes & Noble), which is going to be phased out as states—and the feds—crack down on gift card expiration dates.
Beyond those specific retailers, Henson said, the most popular gift card categories are restaurants and department stores, while the least favorites are toy stores and bookstores. Why are toy store gift cards so unpopular? "Children can't use a gift card," Henson said, and adults are hesitant to give toy store gift cards to other adults, even adults who need to purchase a lot of toys.
Today's larger retailers are experimenting with a wide range of different uses for gift cards, with Subway experimenting with merging loyalty cards, gift cards and payment cards.
Security issues are another concern, with e-commerce gift cards posing particular security and fraud challenges, and some retailers are experimenting with adding a secondary identification number on the card, à la traditional payment cards.
One potential advantage for retailers working with Plastic Jungle is a wealth of CRM (customer relationship management) data. Today, gift cards are plastic marketing frustration because the person who buys the card is not the person who will use it, so correlating usage data with the purchaser's identity is worthless.
"Right now, retailers are selling blind because the whole product is designed to be given away," Henson said. "We're going to be able to gather that data and see who is actually using the cards and—more importantly—who doesn't want that card."
Initial data to be collected includes age, income level, gender, name, address and related demographic data, she said, but other data points—such as asking why consumers are turning in a particular card—could be added shortly.
The company, which now employs "fewer than five" full-time workers, is also making a plea for charitable contributions with small dollar amounts left on gift cards, Henson said. "Often, people will use almost their entire gift card amount at a store, but then have a little money left on the card," she said. "These low-balance cards typically get forgotten and go unused, but now people can donate them to worthy causes. All those less-than-$5 cards can add up to a great deal of help."
Trying to address the hundreds of millions of dollars lost each year in gift cards that are never redeemed, Plastic Jungle, of Fresno, Calif., has created a Web gift card exchange with a twist: The company is selling valid gift cards for less than the card's face value, which could have a significant impact on retail gift card sales.
Plastic Jungle CEO Tina Henson said the company plans on partnering with major retailers, to help them reduce accounting losses from unredeemed gift cards.
"Home Depot had $43 million gift cards that were unredeemed and more than two years old," said Henson. "It just goes on their books as a liability because they're not able to show it as revenue until it's been redeemed."
The way it works is consumers go to the site and fill out some forms (the current site doesn't ask much but that's going to change by mid-October), and they can then sell whatever gift cards they have for something in the range of 65 to 75 percent of the card's remaining value. The site charges a flat fee of $3.99 for each gift card listed, regardless of value, and all transactions are processed through PayPal.
Consumers can also purchase gift cards and pay about 90 to 95 percent of the card's remaining value, with PlasticJungle.com keeping the difference, Henson said. This would allow consumers to purchase, for example, a $100 Home Depot gift card for $90.
The actual percentages paid and received will vary based on supply and demand, Henson said, adding that some gift cards are simply more popular than others and therefore demand a higher percentage. What are the most popular gift cards? Those from Home Depot, Wal-Mart and Target. The least popular? Barnes & Noble and Borders.
Henson speculated on a few reasons for those popularity differences, but thought a key reason are that those booksellers happen to still use expiration dates (two years for Borders and one year for Barnes & Noble), which is going to be phased out as states—and the feds—crack down on gift card expiration dates.
Beyond those specific retailers, Henson said, the most popular gift card categories are restaurants and department stores, while the least favorites are toy stores and bookstores. Why are toy store gift cards so unpopular? "Children can't use a gift card," Henson said, and adults are hesitant to give toy store gift cards to other adults, even adults who need to purchase a lot of toys.
Today's larger retailers are experimenting with a wide range of different uses for gift cards, with Subway experimenting with merging loyalty cards, gift cards and payment cards.
Security issues are another concern, with e-commerce gift cards posing particular security and fraud challenges, and some retailers are experimenting with adding a secondary identification number on the card, à la traditional payment cards.
One potential advantage for retailers working with Plastic Jungle is a wealth of CRM (customer relationship management) data. Today, gift cards are plastic marketing frustration because the person who buys the card is not the person who will use it, so correlating usage data with the purchaser's identity is worthless.
"Right now, retailers are selling blind because the whole product is designed to be given away," Henson said. "We're going to be able to gather that data and see who is actually using the cards and—more importantly—who doesn't want that card."
Initial data to be collected includes age, income level, gender, name, address and related demographic data, she said, but other data points—such as asking why consumers are turning in a particular card—could be added shortly.
The company, which now employs "fewer than five" full-time workers, is also making a plea for charitable contributions with small dollar amounts left on gift cards, Henson said. "Often, people will use almost their entire gift card amount at a store, but then have a little money left on the card," she said. "These low-balance cards typically get forgotten and go unused, but now people can donate them to worthy causes. All those less-than-$5 cards can add up to a great deal of help."
Thursday, September 21, 2006
House Panel Approves Electronic Surveillance Bill
Legislation would grant the administration 'even broader authority to spy on Americans,' say privacy advocates.
Grant Gross, IDG News Service
A U.S. House of Representatives Committee has approved a controversial bill that would broaden the U.S. government's ability to conduct electronic surveillance on U.S. residents by making it easier for federal law enforcement officials to get court-issued warrants.
The Electronic Modernization Surveillance Act, opposed by several privacy groups, would also allow federal law enforcement officials to spy on U.S. residents for up to 90 days without a court order in the period after a terrorist attack. The House Judiciary Committee approved the legislation Wednesday by a 20-16 vote, with all committee Democrats present voting against the bill.
The bill, sponsored by Representative Heather Wilson (R-New Mexico, would reduce the amount of information required from federal agents applying for a wiretapping warrant from the U.S. Foreign Intelligence Surveillance Court. The bill would clarify that the U.S. government can seek wiretaps on any type of electronic communication, not just telephone- or radio-spectrum-based communication.
A Tool Against Terrorism?
Republicans praised the bill, saying it will help the U.S. government fight terrorism. The bill will provide the U.S. intelligence agencies "greater agility and flexibility as they try to thwart our determined and dangerous terrorist enemies," Committee Chairman James Sensenbrenner (R-Wisconsin) said in a statement.
The full House is expected to vote on the bill by the end of the month. The committee's action comes after President George Bush called on Congress to approve a controversial electronic surveillance program conducted by the U.S. National Security Agency. The NSA has conducted the program, reportedly targeting U.S. residents speaking with foreigners who have suspected terrorism connections, without getting court warrants.
Bush has asked Congress to authorize the NSA program to ward off multiple court challenges against it. In August, a U.S. judge in Michigan ruled the NSA program is illegal and must be halted. The Bush administration has appealed that ruling.
Or an Attack on Privacy?
The Senate Judiciary Committee approved three surveillance bills last week. The Center for Democracy and Technology, a privacy and civil liberties advocacy group, called the Wilson bill and the Senate's National Security Surveillance Act two of the worst bills now in Congress.
"Couched in the seemingly laudable terms of 'modernization,' the bills would radically undermine the privacy of innocent Americans -- not just by legitimizing the administration's warrantless surveillance programs -- but by granting this and future administrations even broader authority to spy on Americans in the United States without judicial review," the CDT said on its Web site.
Grant Gross, IDG News Service
A U.S. House of Representatives Committee has approved a controversial bill that would broaden the U.S. government's ability to conduct electronic surveillance on U.S. residents by making it easier for federal law enforcement officials to get court-issued warrants.
The Electronic Modernization Surveillance Act, opposed by several privacy groups, would also allow federal law enforcement officials to spy on U.S. residents for up to 90 days without a court order in the period after a terrorist attack. The House Judiciary Committee approved the legislation Wednesday by a 20-16 vote, with all committee Democrats present voting against the bill.
The bill, sponsored by Representative Heather Wilson (R-New Mexico, would reduce the amount of information required from federal agents applying for a wiretapping warrant from the U.S. Foreign Intelligence Surveillance Court. The bill would clarify that the U.S. government can seek wiretaps on any type of electronic communication, not just telephone- or radio-spectrum-based communication.
A Tool Against Terrorism?
Republicans praised the bill, saying it will help the U.S. government fight terrorism. The bill will provide the U.S. intelligence agencies "greater agility and flexibility as they try to thwart our determined and dangerous terrorist enemies," Committee Chairman James Sensenbrenner (R-Wisconsin) said in a statement.
The full House is expected to vote on the bill by the end of the month. The committee's action comes after President George Bush called on Congress to approve a controversial electronic surveillance program conducted by the U.S. National Security Agency. The NSA has conducted the program, reportedly targeting U.S. residents speaking with foreigners who have suspected terrorism connections, without getting court warrants.
Bush has asked Congress to authorize the NSA program to ward off multiple court challenges against it. In August, a U.S. judge in Michigan ruled the NSA program is illegal and must be halted. The Bush administration has appealed that ruling.
Or an Attack on Privacy?
The Senate Judiciary Committee approved three surveillance bills last week. The Center for Democracy and Technology, a privacy and civil liberties advocacy group, called the Wilson bill and the Senate's National Security Surveillance Act two of the worst bills now in Congress.
"Couched in the seemingly laudable terms of 'modernization,' the bills would radically undermine the privacy of innocent Americans -- not just by legitimizing the administration's warrantless surveillance programs -- but by granting this and future administrations even broader authority to spy on Americans in the United States without judicial review," the CDT said on its Web site.
Subscribe to:
Posts (Atom)