Sunday, July 18, 2004

Collaborative Filtering

Craig Newmark notices something about the list of most-referenced movies and those that are missing from the New York Times list.

One thing I wanted to mention was another way of rating movies: to predict how much you will like a given movie based on your ratings of other movies and comparing that to the rating given by others who liked the same movies you did. The mail-order DVD rental website Netflix includes just such a system. There were a few earlier experiments with this on the Internet, notably EachMovie.com and MovieCritic.com, now both defunct. The only free website I know of that currently rates movies this way is MovieLens.

DEC (Digitial Equipment Corporation, later bought by Compaq which later merged with Hewlett-Packard) sponsored Eachmovie.com and made the data from the project available for research purposes. As part of class project at UC Berkeley, a principal components analysis was done on the data. The results are interesting. The meaning of the first three components they found seem pretty natural, but I'm not sure about how they named the next two:

COMPONENT 1: HIGHBROW to LOWBROW (magnitude = 128)

58 Il Postino (The Postman)
52 Mighty Aphrodite
41 Richard III
232 Eat Drink Man Woman
306 Three Colors: Red
608 Fargo
265 Like Water For Chocolate (Como Agua Para Chocolate)
307 Three Colors: Blue
25 Leaving Las Vegas
111 Taxi Driver (1976)
800 Lone Star
374 Richie Rich
65 Bio-Dome
267 Major Payne
420 Beverly Hills Cop III
174 Jury Duty
135 Down Periscope
810 Kazaam
419 The Beverly Hillbillies
66 Lawnmower Man 2: Jobe's War
502 The Next Karate Kid
688 Operation Dumbo Drop
1250 The Bridge on the River Kwai (1957)
1019 20,000 Leagues Under the Sea (1954)
1262 The Great Escape (1963)
1195 The Magnificent Seven (1954)
1287 Ben-Hur (1959)
1283 High Noon (1952)
914 My Fair Lady (1964)
899 Singing in the Rain (1952)
1231 The Right Stuff (1983)
1028 Mary Poppins (1964)
1012 Old Yeller (1957)
85 Angels & Insects
81 Things to Do in Denver When You're Dead
180 Mallrats
766 I Shot Andy Warhol
444 Even Cowgirls Get the Blues
663 Kids in the Hall: Brain Candy
347 Bitter Moon
778 Trainspotting
352 Crooklyn
259 Kiss of Death
156 Blue in the Face
421 Black Beauty
262 A Little Princess
46 How to Make an American Quilt
261 Little Women
531 The Secret Garden (1993)
238 Far From Home: The Adventures of Yellow Dog
271 Losing Isaiah
270 Love Affair
8 Tom and Huck (Tom Sawyer)
258 A Kid in King Arthur's Court
351 Corrina, Corrina
1476 Private Parts
842 Tales From the Crypt Presents: Bordello of Blood
879 The Relic
1431 Beverly Hills Ninja
1405 Beavis and Butt-head Do America
1407 Scream
798 Daylight
799 The Frighteners
1429 Jackie Chan's First Strike
1004 The Glimmer Man
1120 The People vs. Larry Flynt
COMPONENT 4: CHICK FLICKS to HORROR (magnitude = 26)
1353 The Mirror Has Two Faces
1363 The Preacher's Wife
1043 To Gillian on Her 37th Birthday
605 One Fine Day
830 The First Wives Club
613 Jane Eyre
838 Emma
1367 101 Dalmatians (1996)
1113 The Associate
848 The Spitfire Grill
17 Sense and Sensibility
328 Tales From the Crypt Presents: Demon Knight
426 Body Snatchers
366 Wes Craven's New Nightmare
610 Heavy Metal (1981)
1261 Evil Dead II (1987)
1347 A Nightmare on Elm Street (1984)
332 Village of the Damned
606 Candyman: Farewell to the Flesh
177 Lord of Illusions
188 The Prophecy (God's Army)
413 Airheads
517 Rising Sun
805 A Time to Kill
832 Ransom
161 Crimson Tide
257 Just Cause
490 Malice
544 Striking Distance
423 Blown Away
1092 Basic Instinct
379 Timecop
280 Murder in the First
99 Heidi Fleiss: Hollywood Madam
818 A Very Brady Sequel
312 Stuart Saves His Family
631 All Dogs Go to Heaven 2
343 The Baby-Sitters Club
609 Homeward Bound II: Lost in San Francisco
181 Mighty Morphin Power Rangers: The Movie
711 Flipper
126 The NeverEnding Story III
810 Kazaam
63 Don't Be a Menace to South Central While Drinking Your Juice in the Hood

From The Furrygoat Experience:
Check out a totally cool spiderman movie “The Peril Of Doc Ock”, completely done in Lego stop motion. Check it out. Very neat.


This page is powered by Blogger. Isn't yours?