Liveblogged by LizHenry. Please go in and fill stuff in or make comments.
Moderated by Rashmi.
tags are overloaded
tags on flickr. interestingness. people want an easy way to go back and find content from a previous period. recommendation engine on delicious.
once you've been using a service long enough you have too many tags. something intersting about clusters. clusters are catergories. soon someone will bring hierarchical structure to this
mary hodder: (napserization.com)
rashmi: what are the structures, what are we leading to?
introductions
shawn: people communication over physical locations, how can tagging be combined with geography and communication.
micah alpern. at UA user created data, semi structured data. thoughts about delicious and tags, organizing one's own information. finding other people's content.
mary hodder: figuring out clusters . conversational communities cross-blogs. using tags for maybe 5% of information but not for all info on conversational communities. topic communities. how to weight conversations within those communities.
Liz asks a question.. er
kaliya. identitywoman. tagging, bloggging, identity, going in between different services and making tags that are linked to you as an identity.
mary : users owning their own data. you guys all own a copy of my information and I own a copy too. user must assert ownership and must be able to attach an identity to all that stuff.
kevin burton. rojo . tag infrastructure. feedblog.org. fixing tags. big problem that if i go to blog i can only tag my own post.
Liz mentions shadows. tag and comment up any web page. parallel web.
ryan from technorati. working on grad degree usf. Also works on http://microformats.org Theflickr infers structure out of taggs by clustering them. An addition to intro material should be related tags. it's a crude method– co-citation. it works fairly well for navigation.
Inferring structure from tags will likely building a web rather than a tree– not a hierarchy.
Kaliya: polyarchy?
Kevin b: ooo! i didn't realize it was that kind of conversation! we're gettin' cuddly in here!
Ryan: as we start clustering things we could infer a hierarchy. ontologies.
jonathan boutell
ted rheingold dogster.
drew. bittorrent. looking to use tags on our search engine.
rashmi: you need a layer on top of the tags to navigate the tags. sophisticated ways of navigating tags. recommendations. related. clusters. co-citation. collaborative filtering and recommendation. what will the new algorithms be?
hierarchy. you can't escape. try to explain the animal kingdom to a child using tags. (wow!)
micah: if i'm using someone else's links to see their perspective i can see their different path. can i shift my perspective in the cloud/cluster. multidimensional space.
ryan: about taxonomy. how you make the animal kingdom, plant kingdoms and their hierarcheies without building from the bottom up. is it from looking at what characteristitc they share in common.
liz: actually that is a huge controversy in biology and the cladistics people are at the throats of the other people etc. (link to this)
ryan: the ancient greeks didn't just say there's 5 things etc. and then look for the things that fit. (well, i think they did — Liz) (of course not, but we need to distinguish between how hierarchies are used and how they are built -RyanKing)
rashmi: in anthropology... you can try to model knowledge in diff. cultures. and then once information has a certain depth and breadth...
ryan: there's things we can learn from how the hierarchies are built.
mary: you can make a comparison between what kids know when thy'er 5 years old what do they know about specificity. they know something's a chair but not an armchair or a chesterfield chair. classification, an inbetween level between tagging and hierarchy you can work from this mid-level.
rashmi: areyou saying that tags work at that level?
mary: don't worry if it's an armchair or a stool. how far are you going from the basic category, how many steps away are you when you decide it's a stool vs. a couch or a wing chair. this is a good starting point for building a classification system.
Micah: We ask sellers for 3 diff kinds of info. categories, a top level. what this thing is. atrributes: more multifaceted. description: completely freeform. we can do more to leverage that freeform description. for buyers we're surfacing the multiple facets... you could consider using tags in a similar way. to help narrow search process.
mary: you could build something where tags highly used are the facets but you could end up with the dewey decimal system. unless you intentionally went out looking for stuff to fit everything that isn't popular.
rashmi: ......
micah....
ryan: .... co-citation... related tags... (I missed it)
rashmi explains flickr clusters. who has used them? only a few people.. they're computed on the fly. Do you understand the clusters? what is the cluster about? you sould be able to predict what's there when you go into a cluster.
mary: it depends on what your goal is and topic is. this comes up. when i see any large body of tags. sometimes i really want all the tagged stuff and they're sending me to this other place. it's not always useful. If i don't know what i'm looking for and want to explore, it's good. if i want all the photos from barcamp i don't want clusters. i just want all of it!
ted: entertainment vs information
liz: dictionary vs. thesaurus. you can browse both diff. ways.
....
People are not using this like google.
Kevin: document clustering is much harder than tagging and photos.
Ryan: photos are compelling. there is something concrete there.
kevin: do people think tags are taking off b/c they haven't hit mainstream yet?
tags are working b/c people aren't trying to game it, break it, spam it.
everyone agrees ruefully
micah: thats why we need identity!!!!!
jonathan: we can tag things ajax, b/c we have a common understanding of the term
-ryan points out that flickr effectively custers the severarl meanings of 'ajax'
I go on for a sec about dictionaries and multiple definitions of the same word and the other info in a dictionary which is not equally weighted tags but info in diff. categories.
ryan: spam and tagging. it's heating up.
rashmi: delcious and identity.
mary: walled gardens. protect your identity and what you've created. you've got a long list of urls you've created. if i started spamming
kevin burton suggests that you can't tag until you have reputation.
liz: there are huge problems with that.
mary: problems!!!
kevin: tough, you ahve to work harder. like on ebay w/out repuation you have to sell cheaper.
micah: the consequences are you end up talking to yoeurself
kevin: people using browsers , if they all start tagging then it's going to suck.
liz: well thats what everyone said about the entire net and then the web.
drue: ....
kevin: .... results too static, top results too static...
ted: did you try saying show me different, show reverse chronology
kevin: using repuation to put in tags.
druie: we need repuation or tags will be meaningless. people can put in their tags.
rashmi: do people agree witih this?
liz: No not at all.
ryan: we can fight spam now because it's dumb.
Liz: when i did web spamhunting for search engine... yes they were dumb... but it needed work...
mary: is there an organization central to fighting spam?
ryan: not quite... (did he say an organization? I missed it) (no organization, but cooperation -RyanKing)
ryan: it's a competitive advantage to fight spam. we're workign with pingomatic...
mary: It's lame to compete on spam control. (hahaha!)
kevin: this is the tragedy of the commons. there will always be parasites. yahoo...
mary: i'm not asking yahoo to change this.
kaliya: ???
mary: why shouldn't we share that info.
ryans: blacklists. ip numbers. realtime blackhole list. when they're wrong, when it's malicious.
mary: why not have a consortium. how do people who have companies get in on that.
ryan : there is no centralized place now. we find webspam with googleadwords, we report it to google. no need for us to tell pubsub or feedster. it benefits everyone in the ecosystem, because the spammers get cut off at the source. we tell the blogspot a whole regexp of domain names is bogus. (yes... I also found htis to be true when spamhunting at excite search). we call rackspace and tell them a whole range of addresses...
kaliya: single signon. can you start inviting people to start using that identity along with tags. follow them.
mary: you mean tag in a distributed way. potentially not having to expose the tag on top of your site the same way you don't have to expose the category of the b log post. they refuse to use tags until they have that option. but they want to be able to make a trusted tag.
rashmi: greater weight will be given to these people's tags...
mary: it still we be underneath the post... there in html but not there in the blog post. whether or not i expose the category is my choice, it's in the rss feed and gets spidered but not viewable to readers of the post. explicit tags . they want catgories or tags on the back end.
rashmi: how does trust/identity mean for that? as a user i go to delicious or technorati. do i have more weight to different people? (do I give more weight to people i choose?)
mary: technorati is set up to recognize if there are lots of links in a post we're going to flag that as potential spam.
(identity... multiple identities.)
jonathan: what about community tagging.
mary: tags as a kind of topic.
multiple social gestures could get you into a community. diff. ways of expressing it. doyou tag yourself? not necessarily. the community is defined various ways
there's overlap between this and the same issues occuring in that other space. (of community and clustering)
by social gesture, Mary means commenting, referring to your name, linnking, the conversant community around you can arise even if you don't have a web presence.
ryan: is using people's proper names useful on the web? I'm one of hundreds of ryan kings.
kaliya: i'm the only one of me....
ryan: but clusters...
.....!
