Wednesday, November 15, 2006

UnSuggester fallout

Our new anti-recommendation engine UnSuggester (blogged about below), has taken off in a weird way. Today was covered in everything from top-shelf library bloggers Tom Roper, Steven Abram and Karen Schneider (via ALA TechSource), to the bouncy, zippy, web-culture vlog MobuzzTV.

Here Mobuzzer Karina Stenquist explains the disconnect between Middlesex and The Passion of Jesus Christ: Fifty reasons he came to die. Karina did something on us before and—I'm sorry—she's great.



I love everything people are saying about it, with big discussions here and on various blogs, mosty people strive to find the oddest, funniest unsuggestions. My favorite blog discussion came off a short note by "Fontana Labs" (a philosophy professor?) on Unfogged, and is now past 120 comments. Much of it came from the proposal:
"There's got to be some ultimately evil book out that that will generate the ideal library. Not The Bridges of Madison County, but like that."
Speculation ensues, with the unsuggestions for Who moved my Cheese? much praised. I enjoyed this exchange:
"Sadly for Unfogged (but perhaps good for Western civilization) Who Moved My Cheese? For Kids, as perfectly soul-deadening a title as I've ever encountered, isn't owned by enough LibraryThing users to produce results. ..."
"Who Moved My Cheese? For Kids. Oh my God, my soul just broke into three thousand tiny wretched pieces."
Hey, I didn't say it. De gustibus...

27 Comments:

Anonymous Anonymous said...

Oddest, funniest UnSuggestion I found was this:

Pump in Alex Comfort's "The Joy of Sex", and most of the top 10 contains books by Chuck Palahniuk.

11/15/2006 4:15 PM  
Anonymous Anonymous said...

Even better - "A Heartbreaking Work of Staggering Genius" returns a book called "Don't Waste Your Life" as the top UnSuggestion. No idea what it's about, but such a fitting title. :)

11/15/2006 4:25 PM  
Anonymous Anonymous said...

I do wish you'd reduce the number of copies required before unsuggestions are generated. I understand you don't want it too low, but how about 50?

11/15/2006 4:30 PM  
Anonymous Anonymous said...

You also got a suggestion on one of the Business 2.0 blogs:

http://blogs.business2.com/softgadgets/2006/11/library_things_.html

11/15/2006 5:59 PM  
Anonymous Anonymous said...

Oops, I meant "mention" not "suggestion".

11/15/2006 6:00 PM  
Anonymous Anonymous said...

So, I looked at unsuggestions for Rita Mae Brown's Bingo, and there was The Little Prince at #39. "But I own that!", I said to myself, "why does it say 'expected 4.7, found 0'?"

It's because 490 people have the book listed without an author!!

(At #5 we find The Screwtape letters ; with, Screwtape proposes a toast, but I own each of those, just separately, not in one volume.)

On the other hand, Chuck Palahniuk and The Purpose-Driven Life show up, as they do with just about every book of mine I've tried here!

11/15/2006 11:48 PM  
Blogger Tim said...

The Little Prince thing is odd. There must be some odd Amazon data in the mix. For that and the other one, I need to facilitate work-combinations across author boundaries.

The Screwtape example brings up that our combining process should have more options, like "part of a series," "contains," etc. It's all planned, but not coming down the pike tomorrow.

11/15/2006 11:52 PM  
Anonymous Anonymous said...

Can you point me to an explanation about what "expected 15.5, found 0" means? Thank you :-)

11/16/2006 10:57 AM  
Blogger Tim said...

Well, it means that given the pool of works owned by people who own X, Y copies of the work could be expected.

For example, imagine a book is owned by 2000 people who have a total of 720,000 books. That's 10% of the books in the system overall (7.2 million). All things being equal that 720,000 books should contain 10% of the copies of every work in the system. So, for example if 1,000 copies of a novel are held overall, one would "expect" 100 in the pool.

Does that make sense?

11/16/2006 11:51 AM  
Anonymous Anonymous said...

Yes, thank you :-)

(The bigger the gap between 'expected' and 'found', the higher up on the UnSuggester list the title appears).

11/16/2006 12:22 PM  
Blogger John Emerson said...

The anti "Wuthering Heights" is half Christian, half programming and AI.

The anti- "Poor Mouth" (Flann O'Brien) includes "Atlas Shrugged, Friedman's "The World is Flat", and "The Purpose Driven Life" in the top 8. This makes it the best book in the world.

But who would have expected the Harry Potter Box Set ne the anti-"Lake Wobegon Days".

I've got more here:

http://www.idiocentrism.com/unsuggest.htm

11/16/2006 1:06 PM  
Anonymous Anonymous said...

I've taken a brief, more personal look at the UnSuggester than has idiocentrism - though some similar results are reached.

11/16/2006 1:35 PM  
Blogger Tim said...

Something is wrong with the HP box set. I think it was a combination thing, happening DURING the query running.

11/16/2006 2:15 PM  
Blogger Tim said...

John: Love your post, but what's up with your comments? Troll-alert!

11/16/2006 2:19 PM  
Blogger John Emerson said...

Tim: Dr. Meat and I get along fine, but I wish he hadn't chosen that thread. It's an ongoing thing.

11/16/2006 2:39 PM  
Blogger Tim said...

Oh. Well, online communication lacks for context, I suppose ;)

11/16/2006 7:44 PM  
Anonymous Anonymous said...

Many of the modern classics taught in public schools (Catcher in the Rye, To Kill a Mockingbird, The Great Gatsby, 1984) feature Christian texts in their negations. What could this mean?

11/17/2006 2:26 AM  
Blogger John Emerson said...

Besides sex, drugs, fantasy, horror, romance and crime, Christians seem to avoid cynicism and social criticism.

By and large, though, when I look up serious non-fiction, the anti-books are pulp and chick lit (together with male chick lit).

"The Great Gatsby" and "To Kill a Mocjingbird" are required reading in many public secular high schools and I bet that's a big factor.

11/17/2006 7:48 AM  
Blogger John Emerson said...

As you said.

11/17/2006 7:48 AM  
Blogger John Emerson said...

The best selector for "serious lit" would be the "Left Behind" series, because it overlaps with Christian literature, pulp fiction (which most Christians don't read) and apparently even chick lit and male lifestyle books.

The "Left Behind" anti-books are like a liberal arts syllabus. The "Left Behind" pro-books list always starts with 2-4 other "Left Behind" books, and then goes on to other very similiar books by Frank Peretti and Jerry Jenkins. A small amount of more serious Christian literature such as Billy Graham's autobography or "The Purpose Driven Life" creeps into the top 20.

11/17/2006 8:10 AM  
Blogger John Emerson said...

The anti-Dean Koontz is also good.

11/17/2006 8:11 AM  
Anonymous Anonymous said...

We need a name for this "male chicklit" - I'd like to suggest either "brolit" or, perhaps more precisely, "brofic"

Thoughts?

11/17/2006 9:57 AM  
Blogger Robert J. said...

In running a handful of my science books through the unsuggester (there aren't all that many that exist in enough copies for it to work) it's rather striking how many times the words "girl" and "sister" and "diary" appear in the titles of the unsuggestions.

11/17/2006 1:37 PM  
Blogger John Emerson said...

I thought "dorklit".

11/17/2006 2:13 PM  
Anonymous Anonymous said...

The programming book "ANSI Common Lisp" turns up quite a few good novels. If I were a programmer I'd be offended at the computer's assumption that I had never read "the Wind in the willows" as a kid.
- tripleblessings

11/17/2006 2:26 PM  
Anonymous Anonymous said...

*anxiously awaits the ThingUnsuggest API*

11/17/2006 6:49 PM  
Blogger Tim said...

Re: *anxiously awaits the ThingUnsuggest API. I can do one, and it's not like we'd be sacrificing *valuable* data. Does anyone really want it?

11/18/2006 12:22 PM  

Post a Comment

<< Home