Technology postsSunday November 06, 2011
WANTED: Algorithmic Help for Netflix
I was doing research on “The Wizard of Oz” recently when I came across this latest horrific example of Netflix's recommendation algorithm:
If you'd asked me to name a thousand movies that were similar to “The Wizard of Oz,” I wouldn't have named any of these. Who is screwing things up at Netflix? Who was fired, and who was hired, and who is running things into the ground there?
As an experiment, I tried it at IMDb.com. These are the recommendations for their “Wizard of Oz” page:
That's a little more like it.
This showed up in my in-box the other day:
Don't quite get the repetition of “Unbroken” but at least it's correct. And late to the game. I read it earlier this year and recommend it. The others either don't appeal (audio Jackie Kennedy) or only appeal if I want to keep up with the conversations on Andrew Sullivan's blog (“The Rogue”).
But at least Amazon's first row of suggetions made more sense than its second row of suggestions:
“Harry Potter” I haven't seen. “Captain America” made my top 10 for the first eight months of the year (at no. 10). “Fast Five” I dismissed in April. And “Transformers”? I think of that franchise as so stupid and noisy as to be part of the general decline and fall of western civilization.
Someone, in other words, needs to work on their algorithms. Or mine.
When I interviewed Jeff Bezos in October 1996, a lifetime ago, he talked about the things Amazon was working on:
Bezos: We want to set the store up so we can redecorate the store for each individual who walks in...
Me: What do you mean?
Bezos: The whole page would be personally designed for you. So if you said, “I really love literary fiction,” well, here’s a great science-ficton novel that we actually think you'd like based on your preferences in literary fiction. Stuff like that. In case you want to broaden out.
It would all be done automatically. It would have to be. The way it might work is you might come in and we present you with a list of 100 books that are in a particular genre, like literary fiction, let’s say, and you would rate the ones you liked the most and disliked the most, and based on what you liked and disliked the computer would be able to form a profile of your particular tastes, and it might try to match you up with people of similar tastes. You call that your affinity group. What are things you haven’t read that people in your affinity group love? And then it would recommend those things to you.
Still a few bugs in the system.
When Writers of Code Write Copy
Not to tread on the territory of AKC, the Copy Curmudgeon, but I saw this the other day while browsing Netflix's site:
It's less what they recommend than how they recommend it.
Based upon my interest in “Straight Time,” starring Dustin Hoffman, and “The Apprenticeship of Duddy Kravitz,” starring Richard Dreyfuss, Netflix automatically generates the following recommendation: Understated Movies based on a book from the 1970s.
By which it means: Understated 1970s movies based on books. (Only one of the above, “Straight Time,” was based on a book from the 1970s.)
Raise the rates, fire the writers, apparently.
Best at Something: My No. 1 Google Rankings
Several years ago my friend Mike told me he was going to be Best Man at an upcoming wedding. He paused a moment before adding, “It's nice to be best at something.”
In that spirit, here are a few of the Google searches where, according to statcounter.com, I have the No. 1 ranking. I've included links to where those searches lead. Your results may vary. Have at:
- dialog between preacher meacham and doc earn God's presence cowboys and aliens = this review
- HDNet what lead your brother to become a suicide bomber = this blog post
- ivan vasilevich menyaet professiyu vladimer visotski = this blog post
- certified copy, the whole purpose of life is to = this review
- kimane maruge letter from president compensation = this review
- angel toves = another review
- f troop is fuzzy = yet another review
- valid reasons why the yankees suck = this article
- yankee fans batteries mariners = yep, this same article
- average weekly movie attendance = post
- midnight in paris quotes zelda fitzgerald you have a stuned look in your eyes, you look anesthetized, lobotomized = category of posts
- I felt the death loneliness that comes at the end of every day that is wasted in your life." = here
- girl who played with fire how does lisbeth escape buried alive = here
- Frank Crosetti Ball Four = here
- johnny damon 3000 = here
- Erik Lundegaard = whew
“Dad, 'F Troop' is fuzzy.” A serious man, making the world safe for unserious TV.
What Bing Does Better Than Google
In my review of Woody Allen's “Midnight in Paris,” which, to me, is about traveling to a place and time where art and literature matter, I made a passing reference to the fact that even Philip Roth doesn't read fiction anymore. He said so in a recent interview with Jan Dalley, “Life After 'Nemesis,'” in Financial Times. I wanted to provide a link to that interview so I Googled two terms: “Philip Roth” and “stopped reading fiction.” Here are the top results for that search:
- The New York Times artsbeat blog referencing and linking to Dalley's FT article.
- The Atlantic referencing and linking to Dalley's FT article.
- Althouse referencing and linking to Dalley's FT article.
- Salon referencing and linking to Dalley's FT article.
- Yahoo News referencing and linking to Dalley's FT article.
- James Russell Ament referencing and linking to the Salon article, which references and links to Dalley's FT article.
- Slate republishing the entire FT article with Dalley's byline.
- ArtsJournal referencing and linking to the Slate article.
- The original article.
Ninth. Above it, you have all the sites feeding off it. In the Google world, in the tech world, there's little respect for original content. To me, this lack of respect filters down and you wind up with some aspect of the shitty culture we have.
As an experiment, I cleared my cache and tried the same search on Yahoo, expecting similar results. Nope. Much better. The original article appeared third.
Then I cleared the cache again and tried it with Bing. The original article appeared first.
Guess I'll be using Bing more often.
Now if we only stop calling it “content” and get back to calling it “writing,” maybe it'll pay again.