May 27, 2010 3:51 PM   Subscribe

What is the Biblioracle? The Biblioracle is a service in the person of The Morning News contributing writer John Warner, said service being a recommendation for what book you should read next. What do I have to do to receive a recommendation from the Biblioracle? In order to receive a recommendation, simply list the last five books you’ve read. Based on that list, the Biblioracle will tell you what to read next. The Biblioracle is now closed. Due to overwhelming response, the Biblioracle can only respond to requests submitted prior to 3:00 p.m. ET. If you missed your recommendation this time around, watch for the Biblioracle’s next appearance this summer! The time for submission is closed but it's still interesting to see the recommendations that John Warner made.
posted by Fizz (31 comments total) 8 users marked this as a favorite
This would be pretty easy to automate by analyzing everyone's 5 books and generating recommendations purely from that.
posted by mhjb at 3:58 PM on May 27, 2010

Oh Amazon already does something similar but this is based solely on an individual's knowledge, a human.
posted by Fizz at 4:00 PM on May 27, 2010

I wonder if Netflix has thought about licensing their algorithm to book sellers and/or libraries. Clearly there's a huge demand for such a thing.
posted by filthy light thief at 4:02 PM on May 27, 2010 [3 favorites]

This is great, even ideal, except for the laziness factor: "Most of these people probably won't read what I recommend anyway." No way to gauge interest or reaction or reputation of those requesting recommendations. How long can it possibly last before there's an urge to outsource?

Oh, and the "people with lame book lists" factor. "Who are these people, a bunch of white male 18-year-olds?"
posted by circular at 4:04 PM on May 27, 2010

What if I hated the last five books I read?
posted by ODiV at 4:04 PM on May 27, 2010 [2 favorites]

Now, for an unlimited time, you can go to any public library and ask any librarian the same question...and get a recommendation! Ask now, while you still have a public library!
posted by Hildegarde at 4:07 PM on May 27, 2010 [14 favorites]

I freaking KNOW what book to read next: Any of the several dozen unread books I've already got sitting around my house. If I buy any more books to read "someday" I'll end up on one of those TV shows about hoarders.
posted by LastOfHisKind at 4:13 PM on May 27, 2010 [7 favorites]

Sort of previously?
posted by elizardbits at 4:27 PM on May 27, 2010 [1 favorite]

it's still interesting to see the recommendations that John Warner made.

posted by drjimmy11 at 4:32 PM on May 27, 2010

Oh it's in the comments. which load eventually.
posted by drjimmy11 at 4:33 PM on May 27, 2010

The other stuff he writes is geared towards Funny. So I guess this is as well.

Or maybe not. Difficult to say.
posted by IndigoJones at 4:45 PM on May 27, 2010 has been doing this for a while.
posted by teraflop at 4:50 PM on May 27, 2010 [2 favorites]

Netflix can't really reuse their algorithm that easily in the book market.

I worked for an online audio book company, and one of the main issues I always had with creating a decent book recommendation engine, akin to Netlfix's, is all the extra data points usable for a movie.

Books will have a title, an author, a publisher, editor, and genre classification that's all easy enough to get at.

A movie has a title, screenplay writer, producers, directors, cinematographers, actors and actresses, genre, and well, a lot more random info some of which is specific to genre, (am thinking action sequence choreography). I think things like the people behind the characters often have big influences on how people like movies. If you like George Clooney or whoever, you may like a lot of the other movies he involves himself in. Then the movies where two people the algorithm thinks you may like are in the same movie.

Books we can't do that with. Not easily. Oh, you like Author A? That's great, we'll give you more Author A books till you've read them all! It's harder for us to make lines between the works of two authors strictly using an algorithm. We can make social recommendations, the whole "People who liked this, also liked that!", but we've seen some fun recommendations via Amazon of the type of thing that can lead to.

There isn't a database detailing the types of characters in books that we can use to say, oh, you like your characters to be dark and mysterious, this book here also has characters that are dark and mysterious, or your plots to have these characteristics. I think the same actor in different roles helps bring a bit of sameness to each character they portray. It creates something familiar and comfortable to the viewer that often has them liking a movie purely cause someone they feel strongly about was in it. If people are comfortable with the character in one role, they're likely to like them in another role.

I do think I remember reading, (likely posted here), about an engine that was working on processing the actual text of books to pick out language to suggest themes/character types, but I've never been able to find it again.

Anyhow, just my rant about how hard it is to 'be like Netflix' in suggesting books to people using nothing but basic book data and some code.
posted by nerhael at 4:51 PM on May 27, 2010 [4 favorites]

Er, actually none of the netflix challenge 'winners' have really used any of that extra information, and instead rely (in general) on factor analysis + a grab bag of machine learning techniques on nothing more than titles and ratings. There is no reason you couldn't use the same approach for books, except that there probably isn't the same amount of data available on people's book ratings.
posted by Pyry at 5:14 PM on May 27, 2010 [1 favorite]

I've found that LibraryThing is really great at recommending books. Some of them are automated, some are based on human recommendations. Of course, I've got over 500 books with ratings in there, not just 5...
posted by DU at 5:19 PM on May 27, 2010 [1 favorite]

Let me guess. You give him your 5 books and he tells you to read Bridge of Birds by Barry Hughart
posted by straight at 5:30 PM on May 27, 2010

straight, here are your last five favorited comments:

Holy shit, Jesus was a Mefite

"Do you have beans but fail to overthink them?"
posted to MetaFilter by grubi at 4:12 PM on May 20, 2010 [7 favorites +]

MeFi post: Why you're probably not named Tricia

Thanks for posting this, even if only for leading me to the Many Eyes site.
posted to MetaFilter by sfts2 at 4:57 AM on May 6, 2009 [1 favorite +]

MeFi post: Hello, Dr. Evil here...

"Too much capitalism does not mean too many capitalists, but rather too few."
-GK Chesteron
posted to MetaFilter by The Esteemed Doctor Bunsen Honeydew at 5:06 AM on May 10, 2010 [6 favorites +]

MeFi post: The Year of the Drone

Seriously though, this whole debate is premised on the assumption that there are two categories of Pakistanis, militants and civilians, and that the former should be reduced by attrition and the latter protected as far as possible. So for each attack, and the campaign in general, we just have to run the numbers: this drone strike killed two militants and eight civilians, but it prevented a suicide attack that would have killed fifteen civilians, so on the balance the score is civilians +5,... [more]
posted to MetaFilter by stammer at 2:20 AM on May 6, 2010 [12 favorites +]

MeFi post: A Cinco De Mayo message to Arizona

This could be even better than Black Dynamite.

Also, we've decided around here that "They just fucked with the wrong Mexican" is a whole genre of film. True Lies? When Bill Paxton is hitting on Arnie's wife? He just fucked with the wrong Mexican. Taken? They just fucked with the wrong Mexican.

And "They just fucked with the wrong Mexican" is an utterly awesome genre. Because in it, people fuck with the wrong Mexican.... [more]
posted to MetaFilter by ROU_Xenophobe at 11:52 AM on May 5, 2010 [36 favorites +]

The Favromancer recommends that you begin following grumblebee more closely.
posted by Kwine at 6:09 PM on May 27, 2010 [2 favorites]

It really isn't that difficult for most readers. You simply notice that Mrs. Jones and Mrs. Smith tend to check out the same types of books. If Mrs. Jones says she really liked a book when she returns it, when you next see Mrs. Smith, you ask her if she's read it. If not, you get it for her.

People think you are brilliant and that you've read all the books in your library.

This doesn't work for people (like me) who read cross-genre. We're a lot harder and a lot easier to recommend books to.

Oh, you don't get books from the library? You use the self-checkout machines and never talk to the circ clerks? You pissed off the nice circ clerks arguing over 15 cents in fines? I guess you can stick with stuff online then.
posted by QIbHom at 6:12 PM on May 27, 2010

There are libraries with self-checkout machines??

Anyway, that system also doesn't work if you read any ghettoized genres, such as SF, that typical librarians turn up their noses at (in a professional way, obvs, but still) in my experience.
posted by DU at 6:20 PM on May 27, 2010

I've never found a book recommendation system that wasn't dreadful.

You like the Golden Compass? Well, then you will LOVE Harry Potter!

You like Peter Straub? Then you will LOVE James Herbert!

Maybe this is only a problem with fans of genre fiction, I don't know, but the automatic assumption seems to be that fantasy fans indiscriminately like all fantasy, that science fiction fans indiscriminately like all science fiction, and that mystery fans indiscriminately like all mysteries.

For some of us, that might be true, but these recommendation systems seem unable to take into account WHY readers like certain authors or certain books within a given genre -- we all apparently like Tolkien if we like Terry Brooks.

Has anyone discovered a book recommended system that takes into acount things like mood, theme, plot-driven vs. character driven, first-person vs. second or third, richness of vocabulary, denseness of prose, whether that prose is plain or ornate, etc.?
posted by Chasuk at 8:33 PM on May 27, 2010

You like the Golden Compass? Well, then you will LOVE Harry Potter!

I think that is more like -- "You like this popular thing? You must also like other popular things!"

Pick something more obscure from Amazon:


The Amber Chronicals

No Harry Potter in sight.
posted by empath at 8:43 PM on May 27, 2010

ugh, chronicles...
posted by empath at 8:43 PM on May 27, 2010

Chasuk is absolutely correct about genre readers. This is even when the reccomendation algorithm correlates sales, instead of just recommending randomly within the genre.

The problem is that there are a number of non-picky genre readers. These are the people who will simply buy and enjoy any book with, for instance, the words "dragon" or "dark" in the title. I know some of these people, and their recommendations are attrocious. Even if you both liked Neuromancer, their recommendation of Hammerjack is, I assure you, not going to result in reading pleasure for you.

These indiscriminate genre readers far outnumber those of us with more discerning taste. And they gum up the works of algorithms, while simultaneously causing bookstores to devote more shelf space to World of Warcraft novels.
posted by Netzapper at 8:59 PM on May 27, 2010

DU: "I've found that LibraryThing is really great at recommending books. Some of them are automated, some are based on human recommendations. Of course, I've got over 500 books with ratings in there, not just 5..."

Came in thread to mention this. Besides showing you human recommendations and machine recommendations, LibraryThing also shows you libraries similar to yours that you can browse at whim. I'd hypothesize their data set is much larger too.
posted by l33tpolicywonk at 9:52 PM on May 27, 2010

Pick something more obscure from Amazon:

Say: The Amber Chronicals

Obscure? OBSCURE?

(stabs self in face)
posted by Justinian at 10:22 PM on May 27, 2010

More obscure than Harry fucking Potter, surely.
posted by empath at 10:35 PM on May 27, 2010

Has anyone seen a Recommend-o-Matic for poetry?
posted by pracowity at 10:53 PM on May 27, 2010

This was fun, even though his recommendation to me was to read Murakami, who I enjoy but don't really get much lasting benefit from. But a great idea, kudos to TMN. It's fun even just to scroll through the comments and see what everyone's reading.
posted by statolith at 6:17 AM on May 28, 2010

And if memory serves, you can tell LibraryThing that its recommendations were wrong and it will improve. Tim Spaulding is The Man; I loves you, LibraryThing. *hug*
posted by wenestvedt at 7:30 AM on May 28, 2010

LibraryThing seems to be a great social solution to the book recommendation problem. For a technological solution, I think we would need to create something like a "book genome" similar to Pandora that would take into account things like mood, reader age level, diction, characterization, descriptive level/technique, plus quantifiable/readily accessible data and all sorts of other things to give the code multiple axes to hang recommendation on. This is mostly inspired by nerhael's fantastic comment above.

This is obviously a more difficult task than Pandora's similar task for music, because it requires a larger time commitment than listening to a song. I shall demand no royalties for whichever enterprising soul decides to bring Beowulf into existence.
posted by Night_owl at 1:28 PM on May 28, 2010

It was fun, and the book looks good. It was so much better than the amazon recommendation system, though, that I am now avowed to ignore amazon. Their recommendations have always been disappointing.
posted by mitzyjalapeno at 8:09 AM on May 29, 2010

« Older The soldiers and police were going from house to...   |   If the past is a palimpsest, what are we? Newer »

This thread has been archived and is closed to new comments