Monday, February 05, 2007

Ten million books!

On Saturday LibraryThing acquired its ten millionth book. Ten million is a bunch. Ten million means something. LibraryThing is no longer a "worthless jumble" of books and tags. It's, it's...
  • A meaningful jumble of books and tags
  • A hook to hang a bunch of pretty charts and graphs
  • An excuse for a book pile contest
  • A cause for celebration, and a party
  • A special cause for celebration for the guy who added number 10,000,000
  • An occasion to lay out future plans and goals
And probably a few other things. Anyway, it's big enough that it won't fit in one blog post, and with everything we have to do and all the vinho verde we need to drink this week, I'm expecting ten-million blog posts to drag on for days.

A meaningful jumble of books and tags. Ten million books translates into a piles of data, and piles of data means fun with statistics. And we've been having fun.

Today I added a new "combined" recommendation list. It draws on LibraryThing's five existing recommendation algorithms to come up with a "best" list. I've replaced the longer list of recommendations on the work pages, wth a link to the Suggester page, where you can see all the lists.* (I'd be interested to hear if people appreciate the simplification or still want the full lists on the work pages.) Combined recommendations are available for 230,000 works. Because of variable work popularity, this amounts to recommendations for 72% of all the books in people's libraries.

Alongside books, LibraryThing's tags have also been growing. Although we've rarely celebrated milestones, tags are the untold story of LibraryThing. LibraryThing members have added thirteen million of them--an unprecented web of meaning in the book world. Check out a tag like chick lit, cyberpunk or paranormal romance and tell me what you think. I think LibraryThing members have arrived at something close to the paradigmatic reading list for these hard-to-pin-down genres.

On the subjet of tags, I recently did a statistical sample of Amazon's book tagging. I estimate that since November 2005, Amazon customers have added about one-million book tags. When LibraryThing, a niche site, collects 13 times as many book tags as Amazon, one of the top-ten most visited sites, something is up. I'll blog about it soon, but I think the basic answer is clear. Letting people tag "their stuff" works like gangbusters. Asking customers to tag "your stuff" doesn't. People make their beds every day. But nobody goes down to the local Sheraton to fluff their pillows.

*Not quite. There are actually ten recommendation lists at play since, when the recommendations are sparse, we factor in a "flip-around" of the recommender-recommended relationship.

8 Comments:

Blogger AndrewB said...

Congratulations gang, I'm sure I'm not alone in that I've been madly refreshing the Zeitgeist all weekend and today to see the magic number - and you're already past it by 83,000 (well, more by now)! Yay!

2/05/2007 7:15 AM  
Anonymous Anonymous said...

Just one more step towards world domination!

2/05/2007 7:44 AM  
Blogger Dystopos said...

People make their beds every day? I need to get with the program.

2/05/2007 9:49 AM  
Anonymous Anonymous said...

I donot have any idea of what a blog is, how to get one and what to do with it once I do get it. any help would be greatly appreciate. thanks, Lilith
e-mail add: msgrace53@hotmail.com

2/05/2007 12:54 PM  
Anonymous Anonymous said...

If tags are valuable, which I agree that they are, then combinations of tags are even more valuable, since you allow more than one tag per book. In your example, you encouraged people to look up "paranormal romance." But I might tag my book "paranormal" and "romance", but not "paranormal romance."

Searching for a tag is a one-dimensional search, but searching for a combination of tags is multi-dimensional. This is meaningful in that it can span multiple subjects easily.

For example, if I would like to find books on farming and theology (somewhat disparate topics), I can do that within my own library from this page:
http://www.librarything.com/search.php
by searching for "farming ++theology".

But I can't do that for all users (it doesn't work), but I would really like to.

2/05/2007 1:01 PM  
Blogger Tim said...

I agree. Very much so. It's contemplated. The problem is processing. Tag pages are cached now, because they take so much processing to make. I think we're going to be pushed to a summary table: tags->work. If we had that, doing unions would be easy.

2/05/2007 1:19 PM  
Anonymous Anonymous said...

Congrats!

I'd like to see more (possibly all) recommendations on the work page. I'm always in favour of less clicking.

2/05/2007 3:09 PM  
Blogger Unknown said...

Food for thought: Over 10 million served....

And to think that this number does not include the french fries.

Make sure those Servers don't get fried from too many customers.

2/05/2007 8:20 PM  

Post a Comment

<< Home