Joho the Blog
|
|
|
July 25, 2007
Tim Spalding at LibraryThing.com has introduced a new wrinkle in the tagosphere...and wrinkles are welcome because they pucker space in semantically interesting ways. (Block that metaphor!) At LibraryThing, people list their books. And, of course, we tag 'em up good. For example, "Freakonomics" has 993 unique tags (ignoring case differences), and 8,760 total tags. Now, tags are of course useful. But so are subject headings. So, Tim has come up with a clever way of deriving subject headings bottom up. He's introduced "tagmashes," which are (in essence) searches on two or more tags. So, you could ask to see all the books tagged "france" and "wwii." But the fact that you're asking for that particular conjunction of tags indicates that those tags go together, at least in your mind and at least at this moment. Library turns that tagmash into a page with a persistent URL. The page presents a de-duped list of the results, ordered by interestinginess, and with other tagmashes suggested, all based on the magic of statistics. Over time, a large, relatively flat set of subject headings may emerge, which, subject to further analysis, could get clumpier and clumpier with meaning. You may be asking yourself how this differs from saved searches. I asked Tim. He explained that while the system does a search when you ask for a new tagmash, it presents the tagmash as if it were a topic, not a search. For one thing, lists of search results generally don't have persistent URLs. More important, to the user, tagmash pages feel like topic pages, not search results pages. And you may also be asking yourself how this differs from a folksonomy. While I'd want to count it as a folksonomic technique, in a traditional folksonomy (oooh, I hope I'm the first to use that phrase!), a computer can notice which terms are used most often, and might even notice some of the relationships among the terms. With tagmashes, the info that this tag is related to that one is gleaned from the fact that a human said that they were related. LibraryThing keeps innovating this way. It's definitely a site to watch. [Tags: tags folksonomy librarything tim_spalding everything_is_miscellaneous] Posted
by D. Weinberger at July 25, 2007 03:00 PM
|
Comments
We have two blogs, so I hope you don't mind me putting the URL here: http://www.librarything.com/thingology/2007/07/tagmash-book-tagging-grows-up.php
Posted by: Tim | July 25, 2007 05:36 PM
When you say that the tagmashes have a persistent URL - does that mean that after creation of the tagmash it does not change? So that when people add new tags to books - still the URL displays the same old list of books? Or does it just mean that the there is a URL that finds all the books tagged with a particular combination of tags? If it is the second meaning then the sentence:
"For one thing, lists of search results generally don't have persistent URLs."
is not correct. I don't remember any search engine that does not give you a persistent URL for a query results - they all use GET method for the queries.
Posted by: Zbigniew Lukasiak | July 27, 2007 05:16 AM
Zbigniew, good point. I meant the latter, and you're right that search engines give a persistent url with dynamically generated content, just like the tagmash url. So it is really a matter of positioning and perception. The Google url for a search on france and wwii is:
http://www.google.com/search?source=ig&hl=en&q=france+wwii
LibraryThing's url for the tagmash is
http://www.librarything.com/tag/france,wwii
So, it's more human readable. More important, the tagmash page tries to assemble resources related the tagmash. So it _feels_ more like a permanent destination than a Google results page does, even though you are right that both are dynamically assembled based on info encoded in the url, and the url is in both cases therefore persistent.
Unless I'm misunderstanding how it works, of course.
Posted by: David Weinberger | July 28, 2007 12:25 PM
Tim Spaldings once more suprises me with a great implementation of a simple function. From a library and information science point of view Tagmash is little more then saved searches of boolean retrieval in a subject indexing system. But this "little more" makes the difference - like collaborative tagging differs from traditional subject indexing systems: It's the interface, stupid! Tim makes Information Retrieval not suck :-)
Posted by: Jakob Voss | July 29, 2007 06:05 PM
Over on Flickr they call them "Clusters." Been around for a while. Unique URL, topic vs saved search presentation, etc.
http://www.flickr.com/photos/tags/wwii/clusters/
And del.icio.us intersections.
http://del.icio.us/tag/wwii+france
It's all in the UI, I will give you that. But behind the scenes, nothing new there. It's damn resource intensive to do in a scaleable way is why we don't see it often.
Posted by: Boris Anthony | July 30, 2007 10:21 AM
Late response. I'm not claiming this is curing cancer or, for example, the intersection of cancer and gout. But, on the last comment:
1. It's not like Flickr clusters because the matches are made by humans, not automatically. Flickr often works, but not always. The tag cluster for "LibraryThing," for example, is a mess. The data just won't support it in many cases. In other cases I think that, although there are potentially interesting intersections, they're not necessarily the ones Flickr discovers.
2. They're more like Del.icio.us. They differ in that, after a human finds one, they're saved. You'll see "wwii, france" on the france and wwii tag pages, on relevant Library of Congress Subject Heading pages, on other tag mashes, etc.
The option to look at the intersection "wwii+france" on Del.icio.us doesn't show up on the wwii and france tag pages. The user does the meaning-making, assembling the terms that make sense together. But the meaning-making doesn't spread to others.
Also, for what it's worth, I don't think Del.icio.us allows negative tags.
Posted by: Tim | August 24, 2007 12:16 PM