Monday, February 02, 2009
Unsuggester my eye
Have you heard of the Unsuggester? It works like this: you enter the name of a book you've read or own, and it comes up with a list of books you're least likely to have on your shelves alongside it.
Out of curiosity, I tried my own first novel, breaking the spirit of my 'don't Google for yourself' rule. And I'm a bit fed up.
This is why. The books purportedly least likely to be next to mine are largely books I admire, and include quite a few that I actually do own - and I probably own more copies of Bareback than anyone. (Most of them are in a box under the stairs rather than on my shelves, but I think that should count.)
Let's take the top twenty, and you'll see what I mean.
1. Heart of Darkness by Joseph Conrad
Don't own a copy, but my fiance does, and I've read it. More than once.
2. Love in The Time of Cholera by Gabriel Garcia Marquez
Don't own it, but have read about half of it.
3. Angela's Ashes by Frank McCourt
Own it. Read it at least twice. Like it a lot.
4. Mere Christianity by C. S. Lewis
First fair cop: I haven't read that, nor would I care to. I've read a few of his essays here and there, though.
5. Atonement by Ian McEwan
Don't own it, but have read it.
6. Dress Your Family in Corduroy and Denim by David Sedaris
Own it. Read it. Love it. I've got almost all David Sedaris's books, including one I queued for him to sign.
7. The Trial by Franz Kafka
I haven't read it, but I've read some of his short stories.
8. Godel, Escher, Bach: An Eternal Golden Braid by Douglas R. Hofstadter
Second totally fair cop: I haven't heard of this.
9. White Teeth by Zadie Smith
Okay, I haven't read this, but on the other hand I reject this as a fair cop: I've actually been anthologised alongside Zadie Smith, in a published collection of Oxbridge student short stories. Never mind being on the same shelf, I've been in the same volume.
10. Tuesdays with Morrie: An Old Man, a Young Man, and Life's Greatest Lesson by Mitch Albom
Fair cop number three.
11. Alice's Adventures in Wonderland by Lewis Carroll
Own it. Own more than one copy, in fact.
12. She's Come Undone by Wally Lamb
Fair cop four.
13. The Secret History by Donna Tartt
Now that really hurts my feelings, as it's one of my favourite books. There are two copies in the house, one pristine copy my fiance owns and one beat-up one that I'm too sentimentally attached to part with. I've got a copy of The Little Friend, too, and I've read that three times.
14. Fear and Loathing in Las Vegas: A Savage Journey to the Heart of the American Dream by Hunter S. Thompson
I used to own a copy of this, though I admit I didn't get round to reading it; it finally fell prey to my 'If you haven't read it within a year of buying it, it goes to a charity shop' rule. But it's been in my library. And I've read Hell's Angels.
15. About a Boy by Nick Hornby
Haven't read it, but I have read other Nick Hornbys, and I own a copy of Fever Pitch.
16. The Virgin Suicides by Jeffrey Eugenides
Fair cop five. I keep meaning to read it, though.
17. Nine Stories by J. D. Salinger
Well, no, but I've read Catcher in the Rye.
18. Choke by Chuck Palahniuk
Fairish cop; I've tried Palahniuk and he wasn't my cup of tea, no disrespect to the man. I'll give that half a point.
19. White Noise by Don DeLillo
Fair cop six and a half.
20. Midnight's Children by Salman Rushdie
Don't own it, but have read it.
I'm stretching a point in cases where I've read other books by the same author, and if I didn't do that my list of fair cops would be longer than six and a half out of twenty - but considering that these are supposed to be the least likely books in the world you'd find me near, even with my flexible cop-system I'd expect a higher success rate on their part. I've read at least something by fifteen of the authors on the list.
By way of comparison, take a look at my Suggester list. Its cop-rate is much more like what my Unsuggester list would aspire to. Of the books on Suggester, I haven't heard of most of them; the only one I own is Bitten by Kelley Armstrong, and that's because somebody gave me a copy which I can't give to charity because its first chapter's missing. Given that the main thing in common with the Suggester books Bareback has is that it features werewolves, I rather suspect we're being judged by our covers here.
I don't know how they're calculating this thing, but either I'm being read entirely by people who like my books but don't like any of the books I like, or else there's a kink in the system somewhere. Given my experiences with the Gender Genie, I am growing suspicious of technology. I'm prepared to have a computer program tell me I'm a man, but if it tells me you'll never see my books alongside Donna Tartt or David Sedaris I take exception.
The program, incidentally, is one I heard about in this article from this weekend's Guardian, talking about the concept of 'homophily', or the ignorance-inducing tendency people have to seek out and remain in the company of those who agree with them. Based on the evidence of LibraryThing, frankly I'm not that worried.
I am torn between being amused and offended ("haha-..WHAT?!")by that thing. I first tried Bareback, to see the complete list you got, and while the top twenty didnt have more than 25% I have/have read(or have the something of the same author), it continued with alot of my favourite books.
I then tried several of said favourite books and it kept coming up with a 20/80 mix of other favourites and books about the bible, interpreting the bible and evangelism. Which in all honesty made it 80% right for the ones I tried, but the thing was obviously cheating, just sticking to one topic/genre.
But then, if you had not written your book, would you read it? You've been fairly clear that you don't read science fiction or fantasy. The group of books which you are suggested with is a fairly standard set of modern/urban fantasy novels and is not really a surprising grouping for your book to have been in.
Also, recall this is books people claim to own -- the types of books I own are a specific subset of the types of books I read.
Though I have no statistics to back this up, I would bet that the group of people who use Librarything regularly are not a representative sample.
In any case, all the unsuggester says is that of the 117 people who use librarything and say they own your book, none of them own these 74 books. It's accurate, for librarything users (I am not one): I own one, arguably two of the books on that list (I own 'The Annotated Alice' and 'Godel Escher Bach'), though I have read over 20 of them (on a brief skimming). It can be fun but it's not particularly meaningful, unless you sold only around the 100 books listed on that site.
I tried it using my absolutely-favoritest book of all time, William Gibson's Neuromancer. My returns were almost all evangelical diatribes, most of them by a guy named John Piper (sample title: Reclaiming Christian Manhood: A Response to Evangelical Feminism. Yikes). While it is in fact true that none of these books will find a home on my bookshelf, it felt a bit like cheating.
Oh, and one Mary Higgins Clark novel. Also a fair cop.
Today's word: hameli - A tasty Eastern European breakfast item made with rolled oats, brown sugar, and pork leavings.
I tried Pride and Prejudice, both in the UnSuggester and the Suggester. It felt quite a bit like cheating. Most of the books the UnSuggester came up with were books on coding and other such non-fiction stuff. The Suggester came up with other Jane Austen books (duh), Charlotte and Emily Bronte (pretty obvious also), and modern books based on Pride and Prejudice.
Incidentally, have you read any of those types? Where modern authors write a "sequel" or retelling of a classic like Pride and Prejudice? I've never read any of them, so I have no idea whether they're any good, but it's always seemed incredibly presumptuous, to me.
Verification word is "emovel": a combination movie/novel for the web.
Kara: The one I'm really looking forward to is PRIDE AND PREJUDICE AND ZOMBIES by Seth Grahame-Smith.
Like SNAKES ON A PLANE, the concept sorta sells itself.
(Verification word: "theyer", one who makes a habit of referring to "What *they* say...")
To give a comparison, I read (and love) your book, and I have also read and fairly liked-to-loved ten of the books in your suggester list. So, the books in that list do really appeal to a someone who read your book.
Well, I tried the UnSuggester and *that* was interesting.
I deliberately put in books as diverse as I could think of -- a romantic suspense title, a religious fantasy, and book of straight up social history -- that were in fact three of my favorite titles ever, re-read until they were in tatters.
Every list came up with other books I owned and loved among the top 20.
I think the fatal flaw of their algorithm is that it doesn't acknowledge that people can be *diverse*. Sometimes I'm in the mood for one thing, sometimes another, and I don't care so much about genre or style or subject as that whatever the author does, s/he does WELL.
(Word verification: "flyzedl" -- It goes ZZZZIPPP when it moves, and BOP! when it stops, and VRRRRR when it stands still...)
It looks like it goes just based on data of what people own - so I'd complain about people judging your book by its cover, not the algorithm. Inability to account for diversity, on the other hand, could well be an inherent problem - after all, it's just looking for the ones you're least likely to own alongside it, and what's that going to be affected by? I would think that truly diverse readers are simply going to average out and have little effect on the calculation. But then it's not really so much a problem as it is something that's outside the scope of what it's doing in the first place.
But then, if you had not written your book, would you read it? You've been fairly clear that you don't read science fiction or fantasy.
Actually, that's not quite fair. I have nothing against science fiction and fantasy - I'd list Jonathan Strange and Mr Norrell as another favourite, for example. Given the right recommendation for my book, I would indeed have read it; I'll try more or less anything.
What I've said is that I don't particularly read science fiction and fantasy, which is something different. I read some of it, the same way I read some crime or some chick lit, but in none of these cases am I an expert on or a devotee of the genre in itself; I just happen to like some books that could be classified thus.
The reason I occasionally mention science fiction in particular is simple: it's not the biggest genre on my shelves, but because I wrote a book about werewolves people sometimes assume it must be, and I don't like to give a false impression. If people assumed I must be a major romance or war stories reader, I'd be talking about that instead; it's just a question of what gets assumed. When I mention it on this blog, it's usually for one of two reasons:
1. Someone seems to have assumed I've been influenced by books I haven't read, and I'm trying to set the record straight.
2. I'm asking for recommendations, and know that the Internet generally has a lot of people who would, given a free choice, mostly recommend science fiction or fantasy because that's their favourite genre. If someone's a fan of science fiction or fantasy as a genre, which a lot of people are, it's been my experience that they'll often be looking for slightly different things in a book - say, good world-building or an entertaining relationship with other words of SFF, or other things that are naturally interesting to a fantasy fan but don't particularly hit my yay!-buttons. One of my best friends absolutely loves SFF, for instance, and there are some books we both love because they hit buttons we share, some books she loves and I don't click with because they hit her buttons but not mine (and those buttons of hers are often SFF ones), and some books I love that she doesn't like because they hit my buttons but not hers. Everyone's button pattern is slightly different. If I was asking her to recommend a book, I'd ask for one that fell into the former category, and I figured that saying 'not science fiction, please' would be a short-handed way of expressing that wish. Possibly it was too short-handed.
But as to whether I'd read my own book, that's kind of the point: the thought that someone who had a similar sensibility to mine, and would in result probably like my book, would end up not reading it because it was classed in a genre that appealed to the wrong yay!-buttons - even though those aren't the buttons the book aims to hit - makes me sad. Being so classed makes me worry it'll put off people I'd consider at least a section of my natural audience, which is to say, people who share my tastes. Of course, not everyone who likes my book will like everything I like - that would be silly - but it saddened me a bit to see me presented as opposite from so many books I really like. The fact that this list suggests even I wouldn't read my own book kind of tends towards the conclusion that it's not a very good classification system.
I don't know how it works, of course. I was worrying it was sales-influenced, but I'm getting the impression it's actually a membership recommendations site, in which case it's not representative and there's no point fretting about it.
Incidentally, have you read any of those types? Where modern authors write a "sequel" or retelling of a classic like Pride and Prejudice?
Not my cup of tea. The reason Pride and Prejudice is good is not because it has a heroine called Elizabeth Bennet, it's because it's written by Jane Austen, and a modern 'sequel' cannot supply that. The USP's gone, to use a marketing phrase. I've read books that engaged critically with the works they're spinning off - Wide Sargasso Sea, The Wind Done Gone - but that's a different matter. I think I've checked a couple of the more straight 'sequels' out of the library in my time, mostly from curiosity, but on the whole, nah. I'd rather read something new.
Word: besesonm. A dialectal word denoting a sweeping brush.
I deliberately put in books as diverse as I could think of -- a romantic suspense title, a religious fantasy, and book of straight up social history -- that were in fact three of my favorite titles ever, re-read until they were in tatters.
Ooh, what were they? (If you don't mind telling me.) That sounds like recommendations...
Sometimes I'm in the mood for one thing, sometimes another, and I don't care so much about genre or style or subject as that whatever the author does, s/he does WELL.
I'm with you! (Well, I almost always care about style, but in every other respect, totally with you - and if we're agreeing that it just has to be done well, I assume we're agreeing that the style has to be at least reasonable.) :-)
Ooh, what were they? (If you don't mind telling me.)
Eep! It's sort of like showing pictures of one's college crushes, but sure:
CROOKED HEARTS by Patricia Gaffney -- The only one of her books I've ever liked, a completely over-the-top farce about two criminal tricksters in an Tinseltown version of nineteenth century San Francisco, who scheme and con and deceive each other into falling in love
TILL WE HAVE FACES by C.S. Lewis -- Lewis's most mature and maddening work, a retelling of the myth of Psyche and Cupid from the perspective of her loving, jealous, wise, selfish, heroic, clueless, and very very human elder sister
WILLIAM MARSHALL, FLOWER OF CHIVALRY by Georges Duby
The thirteenth century from the inside out, through the reconstruction of the life of one man, who rose from the son of a humble knight to the most influential advisor of the Plantagenet kings
(Word verification: "fratomm", a rare sexual fetish involving Gummi bears and lederhosen)
Kit, I would take this list with a shaker of salt. I agree that it doesn't take into account diversity. Yes, the Suggester might pull up a bunch of things that would be found on a bookstore shelf next to B/B, and might even be found to cohabitate with B/B in the home of a lover of alt. historical fiction. However, the principle of the Unsuggester strikes me a dubious. I bet the bookshelves of the commentariat here are pretty colorful and wide-ranging. Some algorithm purports to tell you what's *not* on your personal, overstuffed bookshelf? In the words of Napoleon Dynamite's brother Kip, "Like anyone could even know that!" :)
(today's word: docemant - the quality control worker at a poutly farm who fastidiously checks that each carton contains twelve whole and uncracked eggs.)
OK.. I just typed in Godel Escher Bach, as it is a book on your list that I have on my Primary Bookshelf. (I have a system for my books, but do not ask about The System).
It seems that the books I am least likely to read are mostly Christian books. Of the kind we all know and love from LB Fridays! Of the 73, I have two on my shelves and none on my Primary Bookshelf.
That really is a terrible list of unsuggestions, I loved Bareback and I've read/enjoyed just over half of the books listed.
Delurking because I'm a bit of a nerd who sometimes works designing algorithms like this, and I can take some educated guesses as to why it's not working for you. It's not super clear from the UnSuggester website how exactly it works, but if I were to design it this is what I'd do, ranked in order of plausibility:
1. If I had access to all the members' lists of books owned or read, then I'd simply see which books it never occurred with. Then I'd return the most popular books that never occurred on the same shelf as it. This approach is likely to "break" if the book you're asking about is not very common on LibraryThing users bookshelves: being uncommon, it will likely not occur with MANY books, and thus they won't be very representative of much of anything. I'm not saying your book is unpopular; probably most books published in the last five years would have that problem, and if LibraryThing is an American-based site and you did the search on "Bareback" rather than "Benighted", that wouldn't help. Out of curiosity, do you get the same type of results if you do the search on the other name of your novel? If the results are qualitatively the same (i.e., a bunch of books you like) but the specific titles are all different, that's a pretty good indication that the algorithm doesn't have enough info to do anything other than something that is essentially nearly random.
2. Another possibility is that the algorithm works by comparing the words that tend to be in both books. Which means that the algorithm will be kind of dumb, and will rank books about werewolves and the like as being somewhat similar, and books with unique low-frequency words (as are found in Godel Escher Bach, Alice's Adventures in Wonderland, etc) as rather unlike. I doubt this is what their algorithm does, because there are more intelligent ways to do it, but there are a lot of dumb things in the world.
Anyway, #1 is probably what is going on.
And FYI, I think your writing (and imagination) is fantastic, and I've also read most of the things on that list. :)
Word: nonsfi -- the word for people who hate reading Science Fiction with Imagination.
... Ah. I just tried it with both Bareback and Benighted; the same, because it realised they were the same book. So it's not a completely idiotic algorithm, but that doesn't tell us much about how it's working.
Another thing to keep in mind is that, although you may not be a genre reader, many of your readers probably are (just because many SFF readers are). So if your book tends to appear on the bookshelves of SFF readers then it's reasonable to suppose the the UnSuggester could come back with any comprehensive list of non-SFF titles.
Okay, this will be the last one, I promise - but I fiddled a bit more with the site and am now virtually certain that guess #1 is something like how it works, and is the reason that it returns screwy result for your book.
Your book sold 117 copies on their site, and you'll notice that if you do a search for books that have sold fewer than that, as often as not it doesn't return anything; the data are just too sparse.
I did some searches on things that sold around the same number, and in every case the results were much like yours: lots of interesting books from a variety of genres, with no discernable pattern.
Only once you start looking at books that sold at least 500-1000 copies does it start doing anything intelligent, and then it has the problem others have pointed out, which is an overreliance on one genre.
So -- really, don't worry about it. The algorithm is breaking on your book, so it's revealing nothing valid at all.
pogrel: a mongrel who is part pug.
I put down the Iliad and chose the Fagles translation (I own several). It decided the things I am least likely to own are books about knitting. But in fact I'm quite an enthusiastic knitter. It also added a few books on religion.
I then tried Sexism and God-Talk (Ruether) and the algorithm bet (not real money, alas!) that I didn't have Shakespeare's Romeo and Juliet as well as anything by Gaiman OR Pratchett, and it added a bunch of other books that appear on my shelves.
It are an ass.
(Possibly because it sensed I needed to calm down a bit and stop dealing with head-exploders like the Unsuggester, Blogger offered me the quiet, pleasant verification word "rains" meaning, presumably, "fails to be clear." Kind of like UnSuggester.)
I tried "The Princess Bride" and "Holes", two of my favorite books, and got back a whole bunch of Christian texts. I think Forza is right: we need to enter a whole lot more entries for Bareback/Benighted so it can generate a proper search.
Word is "chropo" -- Not enough money for the back re-alignment.
I tried the Unsuggester on Seamus Heaney's Beowulf (over 700 copies in their files) and I got back a list of books and authors that I'd never heard of. So what does that tell me -- not much.
Then I tried T.S. Eliot's Four Quartets and among the returns were several that I've read and liked, but also several entries from the "Left Behind" series. So they got that much right, at least!
It doesn't seems to work at all for children's books. I entered Half Magic and The Stinky Cheese Man (don't know why those were the first two that occurred to me, but so it was) and got back a whole bunch of Unsuggestions of heavy-duty works of religion and philosophy, a few gory-sounding novels, and Barack Obama's Dreams From My Father. Possibly true, but certainly unhelpful.
"Books I've read" and "Books I've liked" and "Books I own" are three distinct concepts, with some overlaps in the actual sets. The whole thing makes me think of those Venn diagrams of intersecting circles. Tiny circles, compared to the vast undefined space of unknown books in which they float.
So who are these LibraryThing members, anyway? A sort of on-line book club, apparently, but who signs up for it? According to their "Groups" list, librarians are by far the most represented (at least among LibraryThing members who join subgroups, I suppose). Followed by Science Fiction Fans, What Are You Reading Now? and Fantasy Fans - which explains why the Suggestions might have tended toward those genres. "Christian" was way down the list, which is probably why all the religious books keep coming up as least-likely Unsuggestions.
And I wonder how many people really take the time to enter their whole collections, or if they bog down after a few favorites or recent purchases.
*pauses to visualize bookshelves lining at least one wall in every room in the house, shakes head*
Alice's Adventures in Wonderland? Doesn't everybody own that?
cographa: a literary genre consisting of books authored by robots
Amaryllis, quite a number of librarians are on LibraryThing (not me) because librarians tend to be the sort of people who want to catalog their own collections of books, and LibraryThing is about the only free and easy software out there for doing that.
In fact, I know of a few (small) libraries that use LibraryThing for their institutional catalog.
(Word verification: "auker", a very very specialized variety of bird-spotter indeed)
Well, that makes sense.
My daughter had an after-school job as a library page for a while. After her first month, she came home and re-organized her own books neatly by topic and author because she couldn't stand to see a bookshelf out of order any more.
It's a source of endless puzzlement to me that people don't alphabetise their books. I mean, how do they find them? I suppose if you've only got about a shelf's worth it's not that big a deal - and if two rounds of househunting in the last five years have taught me anything, it's that not everybody owns books (that plus textured wallpaper was invented by somebody who hates humanity), but my fiance has about three floor-to-ceiling cases' worth and he navigates them entirely by memory. He swears he knows where everything is, but what about when I want to borrow something of his and don't know where to look, eh? And I'm sure I can cite a few times where he didn't know where one of his books was either.
Word: flestre. A sore common to medieval soldiers caused by the chafing of ill-fitting armour.
I bet your fiance DID know where to look, he was just saying he didn't. After all, sharing the rest of your life with someone is one thing, but sharing a bookcase?
As for how to organise a bookcase, how do you do it? Do you alphabetise by title? By author? By subject? I will now look at the Primary Bookcase for an example..
If I had them in alphabetical order, then Godel Escher Bach an investigation into maths and science and music (but mainly maths) would have to go next to either Mason and Dixon by Thomas Pynchon (fictional history) or Ghengis Khan by John Mann.
I think we can all agree that both would be crazy ideas, can't we?
Well, I put in Clive Barker's _Cabal_ and Neil Gaiman/Terry Pratchett's _Good Omens_, and in both cases got back a list of Christian diatribes, evangelism, and biblical study guides. So, on the one hand, yes, granted; but on the other, it feels like cheating and it's not terribly informative, is it?
I'm afraid our bookshelves are a horrible jumble, but that's more a failure of space and Will To Clean than it is an organizational system. On the other hand, it does have some advantages; I keep discovering things that I haven't read in positively ages, which is always pleasant...
Non fiction is grouped by subject (by my own iconoclastic system), fiction by genre and then alphabetically by author.
Anne Fadiman has a wonderful essay on "Marrying Libraries" in her book EX LIBRIS.
(word verification: "gramsabi" -- a standard unit for measuring the heat of horseradish)
As for how to organise a bookcase, how do you do it? Do you alphabetise by title? By author? By subject?
By the author's surname, except in the case of biographies, where they go under the subject's name - hence, for example, my biography of Jane Austen is next to the Austen novels under A rather than under S for Carol Shields, who wrote it. (It's very good, recommended.) That mixes fiction and non-fiction together, but that hardly matters; it's not as if different genres catch fire in the presence of each other. To my mind, the simpler the system the less confusing.
I put 1984 into the UnSuggester, and got 1984 back as the number one book I'm least likely to own. Animal Farm was number 2. Presumably Orwell's readers were so depressed by his vision of the future they foreswore all his other books.
Highly recommend Godel Escher Bach, actually. I found it a surprisingly entertaining read and I really don't understand math at all, and think I may be dysnumeric. It's not meant to be read all in one bite, mind.
Highly recommend Godel Escher Bach, actually. I found it a surprisingly entertaining read and I really don't understand math at all, and think I may be dysnumeric.
If you read it and understood it, doesn't that falsify "I don't really understand math at all"? :)
This is like a zillion years too late, but my god, that suggester is *horrific*. To be honest, though, Bareback took one concept and faked the other way. Really, I'm not sure if there is anything other than Octavia Butler's ouvre that is really comparable to either of your books. Bareback made me think quite fondly of Bloodchild, actually, and chewing the fat about the differences was fun.Post a Comment
July 2006 August 2006 September 2006 October 2006 November 2006 December 2006 January 2007 February 2007 March 2007 April 2007 May 2007 June 2007 July 2007 August 2007 September 2007 October 2007 November 2007 December 2007 January 2008 February 2008 March 2008 April 2008 May 2008 June 2008 July 2008 August 2008 September 2008 October 2008 November 2008 December 2008 January 2009 February 2009 March 2009 April 2009 May 2009 June 2009 July 2009 August 2009 September 2009 October 2009 November 2009 December 2009 January 2010 February 2010 March 2010 April 2010 June 2010 July 2010 August 2010 September 2010 October 2010 November 2010 December 2010 January 2011 February 2011 March 2011 April 2011 May 2011 June 2011 July 2011 August 2011 September 2011 October 2011 November 2011 December 2011 January 2012 February 2012 March 2012 April 2012 May 2012 June 2012 July 2012 August 2012 September 2012 October 2012 November 2012 December 2012 January 2013 February 2013 March 2013 April 2013 May 2013 June 2013 July 2013 August 2013 September 2013 October 2013