Manage your expenses via Email, SMS, Twitter, Voice (Jott: Call and say your expense), IM (Yahoo, AIM, MSN), or Web.
Taste Vectors: Own, Share Your Tastes via XML
Each time you use Amazon it assembles a better profile of your tastes, your habits, your likes and dislikes. Why should they own your profile? I propose you publish your likes and dislikes via a simple XML format (RSS? Atom?), and we build collaborative filtering tools to help you find people who share your tastes and lead you to new things you’d like to taste.
The basic idea is something like this: imagine we had a vector of likes and dislikes for a group of people. For each person, a long array that has a like or dislike for each item. We can use these vectors to calculate similarity between people’s tastes, and to predict items they may like but have not tried. This sort of technology is fairly common now, with Amazon and Tivo being two
well known examples.
Trouble is, Amazon may have a partial picture of what I’m interested in based on my interactions with them, but they’re not sharing it with me. So why don’t I keep and publish my own taste vector? If you do the same, and we can convince many others to do the same, we can get some meaningful info floating out there.
Here’s my basic proposal: let’s come up with a dead-simple XML format for expressing these vectors, and a convention for where they live - say, http://yoursite/tastevector.xml . We need to agree on a couple of things: a consistent way of referring to items (so we both use the same ID for the same item), and a consistent way of expressing like and dislike.
Consistent naming is a hard problem. So let’s punt and go with a good-enough solution. We can use ISBN for books, IMDB ids for movies, Amazon id’s for items carried by Amazon, and we’ll figure out something for the rest.
For expressing like and dislike, we can start with a very simple setup: Like, Dislike, Neutral, and Own, where Own indicates you’ve bought this thing but haven’t bothered to publish an opinion on it.
If you’re a geek like me, we’re pretty much there. But for the rest of the world, we’ll need tools. This can be very simple - just a web page that lets you type in your item and your opinion of it, and spits back a piece of XML, or maybe updates your taste vector for you. This could be a service that stores and serves up your taste vector so you don’t have to worry about it.
But now we’re publishing lots of details about what people like. What about privacy? Well, come up with a fake name or a fake identity for your taste vector if you’re worried about it. And don’t list the items you’d be embarrassed to tell other people about.
With me so far? If we can get enough people to do this, a bunch of tools will emerge to gather intelligence from it and suggest new things for you to try. Believe me, this is data people would kill for - better yet, it’s data marketers pay lots of money for today.
Why do I care? Because my tastes are sometimes outside of the norm, and I want to find the other wierdos to turn me onto new wierd things to try. If you’ve been following the long tail discussions, I may be able to convince you that there are a lot of outside-the-norm people there - in fact, it’s very likely you are one yourself.
Why do I think this may work? Well, it’d take a miracle to get this to really go where it could, but the success of technorati style folksonomies has me thinking miracles are happening these days.
If you’ve read this far you must be interested, so drop a comment in here and we can see if this has any legs…
This is interesting and I’ve had the same idea some time ago. What is needed is a community to get around this.
I think we could count on Lucas Gonze from webjay.org to help. And they have an XML proposal for playlists… you really want to make everything interoperable… so that if I select songs for a playlist, I can export those as part of a tasteXML format.
I like the idea and I’d certainly be willing to publish my own (minus any embarrassing revelations, such as my predilection toward lesbian rock). I think for any but the geekiest among us, some sort of easy front-end editor would be necessary…something similar to RSS generators, or bookmarklets that could automatically add an item from the current page. For example, if I click my bookmarklet while browsing a movie on IMDB or looking at a book on Amazon, it would automatically gather the necessary ISBN or movie ID and either save it locally or post it to a ‘taste submission’ page on your own site. That would update your taste vector to set the book to ‘own’ and the movie to neutral/own for further updates.
If you really want to make this proposal serious, you should at least have SOMETHING at http://www.parand.com/tastevector.xml ……gotta drink your own Kool Aid.
Ok, the bookmarklet idea is what I was thinking of. Here’s a page on the Amazon url format:
http://reviews-zdnet.com.com/AnchorDesk/4520-7298_16-5089045.html
Turns out there is an ASIN that should be easy to parse out.
Daneil, the playlist idea is an excellent one. Is there an ISBN equivalent for music? If not, we’ll have to go with some normalized form of the artist+title or something like that.
It turns out I’m too much of a geek to just create the tastevector.xml file; first I have to create a tool that creates the tastevector.xml file for me….
I’m thinking of doing the Web programming in PHP, for no particularly good reason…
I guess i’m anti-social or something, but i’m looking for a way to opt-out. I mean enough is enough. Personal Taste’s being displayed in jumbled english defined by greater than and less than symbols? I’m personally looking for the cookie atom bomb. Kabooom!! I don’t care about privacy (per se), but I do care how my information is used and who is using it. I got nervous 2 years ago when Amazon created checklists from previous purchases. now tastevector.xml to be worked over by any parser come across.
Shaggy used to say ZOINKS when bad shit was about to happen…
I guess i’m anti-social or something, but i’m looking for a way to opt-out. I mean enough is enough. Personal Taste’s being displayed in jumbled english defined by greater than and less than symbols? I’m personally looking for the cookie atom bomb. Kabooom!! I don’t care about privacy (per se), but I do care how my information is used and who is using it. I got nervous 2 years ago when Amazon created checklists from previous purchases. now tastevector.xml to be worked over by any parser come across.
Shaggy used to say ZOINKS when bad shit was about to happen…
oops, i was changin my email address to my spam account so that when this page gets crawled the mail account I care about doesn’t get ripped and stored in some spammer db.
oh well…
Ok, after the dog food threats, I decided to do something about it.
I haven’t figured out how to do the auto-install of bookmarklets, so create a bookmark on your link bar in firefox, right-click on it and edit the properties, and set the location to:
javascript:location.href=’http://parand.com/tastevector/maketaste.php?user=myid’
Now go to amazon, browse around and find an item you like, and click the bookmark you created. You should see a version -0.1 of tastevector.xml . Note the actual format of the xml is not at all what it should be, but you get an idea.
Actually, for songs, the best identifier would be the URI where you can find the MP3 file.
[...] couple of years ago I wrote about Taste Vectors, did a bit of hacking, and not much [...]