Will you like it?
I added something I've been working off-and-on for about a year*: "Will you like it?" Here's an example, correctly predicting that I will like Nabokov's King, Queen, Knave:
You'll find the section on work pages.
Because it requires a lot of processing, you have to click to get the result. Here it is, correctly predicting that I would not enjoy a popular book about Knitting:
Each assessment has a "certainty" score (eg., "high," "low," etc.) based largely on how popular the book is. You can see the raw scores by hovering over the downward arrow.
How good is it? Meh. It'sokay.
This is a devilishly hard algorithm to get right. I have some ideas for improvement, but it's fundamentally a lark and a conversation piece at present, so I don't want to waste too much time on it.
How it works. In case you're interested, it works completely apart from our book-to-book recommendation system, or the system that aggregates those recommendations into member-specific lists of 1,000 recommended books. Instead, "Will you like it?" works directly from the data, examining the users who have a book and how their books relate to yours.
As such, it isn't very good at sussing out where your tastes differ from those of people who share your books. For example, my large collection of books on Greek history match me up with people who enjoy other ancient history, but I am not that interested in early Republican Rome, no matter what the algorithm thinks.
What's interesting? I'm not going to claim it's perfect, but it's interesting that, to my knowledge, nobody's every tried this before.
I think this is yet another case of Amazon limiting the horizons of what people imagine online, particularly in the online book world. Amazon pioneered book-to-book and user-to-book reviews. The work was groundbreaking but it was also routed in commercial success. User-to-book recommendations drive customers to books they'll like and book-to-book recommendations help them find the perfect book, as well as increase the number of items in each order. Giving people honest assessments of whether they'll like a book is murkier. Does Amazon want to tell a customer they won't enjoy something? And what if they're wrong?
Meanwhile, LibraryThing succeeds by being fun and interesting, not by selling books. It gives us a rare freedom to invent features that don't sell books, like our Unsuggester—what books will you hate?—and now this.
I started a topic to discuss it.
*Don't worry. This didn't distract. I just pushed two combination/separation bug fixes, and Chris and I are hard at work on the catalog, in preparation for some larger changes (ETA: one week?).
Labels: amazon, new features, recommendations
posted by Tim at 12:00 AM
26 Comments:
This comment has been removed by the author.
The other problem Amazon has is that it has a much more random view of my library. LibraryThing is probably the only group that has this level of data.
BTW: I got "Warning: Division by zero in /var/www/html/inc_willilikeit.php on line 97" on Tour de Lovecraft: The Tales, http://www.librarything.com/work/6060504/book/40168412
Actually, a service called Alexandria Digital Literature had this kind of feature. AlexLit wanted to be an eBook dealer, but their recommendation service was what most of the users loved about it. I'm not sure if it's an ongoing concern any more, though.
I do miss Hypatia from Alexandria Digital Literature—I hear they're trying to bring it back. The one way that it differed from Tim's algorithm is that it was based on how users rated the books they chose.
Trollsdotter
"This is a devilishly hard algorithm to get right. I have some ideas for improvement, but it's fundamentally a lark and a conversation piece at present, so I don't want to waste too much time on it."
Why not run a little contest like the Netflix recommendation engine. Post the algorithm code and see if somebody can come up with some thing better. Post the suggested code for everyone to see. Harness the power of the crowd!
Is there any intention of using the star ratings to help inform recommendations or "will you like it?"s?
Speaking of ratings, I'd like to know how much the predictions match up with ratings. I rated a certain biography of Joan of Arc at 4.5 stars, which I'm told, with "medium" certainty that I "will like".
LOL...it can't figure me out AT ALL!
It's been much more accurate for me than the "recommendations" feature.
This looks great.
(And hey, trollsdotter, I thought I was the only one who remembered Hypatia! I do miss that site.)
Can the "Will you like it?" algorithm predict how likely you are to like the "Will you like it?" feature itself?
That'd be cool.
Fun! For books I've already read, I'd say it's about 80% accurate. (It doesn't seem to know about my adventurous side.)
Vikinga
It's been pretty off for me so far on every book I've clicked on that I've already read -- can't say for sure about its predictions for the books I haven't. Perhaps this is because there are a ton of books in my library that I haven't gotten around to reading yet? Either way, utilizing the ratings system would be a huge boost to the accuracy of this feature, I bet.
Thanks for innovating!
I love this feature; it's totally fun. Thank you, Tim!
Alexandria Digital Literature isn't taking any new members, but if you already have a username and password (and you can find them), you can still use Hypatia! http://www.alexlit.com/
It was a great application that suffered from being harnessed to a lackluster business model. I wish the ADL people would team up with LibraryThing....
It would be great if the system could account for star ratings.
As a very simple measure, how about using only books rated with 4 or 5 stars for comparison purposes?
I try to rate all my books as I read them, but find it frustrating that the ratings don't seem to be taken into account by the recommendations.
I believe, if I understand how it works, that it will be accurate (without having to worry about ratings) for people who get rid of books they don't like (and whose most-similar-libraries do the same). For people like me who don't get rid of books, just rate them low and tag them @Discarded, it's not going to be very good - and the same for anyone who has me as a most-similar-library. That's where ratings would help. But it's fun as is.
Fun to play with! Thanks, Tim--So far, after checking about 6 titles, I find that when it's right, it's very right...about 4 out of 6.
A load of fun and very amusing to play with, but I wouldn't trust it enough to actually buy books based on what it says. So far it has gotten most my selections dead wrong. But who cares? Nothing is ever perfect on day one, especially when it comes to individual tastes (which can sometimes change on a day-to-day basis!)
But it's still fun to play with...kind of like a magic 8 ball, only smarter.
This is great fun, and for my selections pretty dead on..... boy, I must be sadly predictable! Thanks for this new feature.
This comment has been removed by the author.
Neat idea. I just started going down books I rated 5 stars.
First two said I'd love the book.
Third said I'd hate the book. A book I gave 5 stars.
"LibraryThing thinks you won't like The Blind Side: Evolution of a Game (certainty: high)"
Is one of those developments that wasn't being ignored for this the long awaited "lists" feature?
Hugely impressed with this. all my 'tests' have been spot on so far.
Fun to play with too. Now i'm going to spend even more time on LT and even less actually reading....
Sweet feature! I especially like that you report both effect size (e.g. "you will probably like it") and signficance (e.g. "highly certain").
I'm currently going through my "wishlist" and "unread" books using "will you like it?" to prioritize them (it would be great if LT could do this sorting for me -- are there any efficiencies available if you know that you're generating the score for multiple works within a single library?)
Fun! Although it told me I "probably wouldn't like" a book I rated 5 stars.
It would be nice if this engine somehow took my book ratings into account, but in the meanwhile it's an interesting (if untrustworthy) tool.
Post a Comment
<< Home