I'm at a Gartner conference right now (more about that later), but one of the keynotes was by James Burke of Connections [1](and 2 and 3) fame. I've been a fan for a long time, so it was worth the conference just to see him. He gave just the sort of entertaining and thought-provoking talk you'd expect from him.
Anyway, seeing him reminded me about his Knowledge Web project. He is running it as a sort of grassroots effort, so basically an open source project. As I get a little older and have kids, this is the sort of open source contribution I can get behind because it's easier to see a real contribution to more people (many of them kids).
Now, a question. They have some key technology guys already involved, but they haven't yet selected a technology for their final implementation. Is this a space Perl or some other open source product could inhabit? They need some good interactive graphing, but James made it very clear that they want it to be very polished and fancy to get kids involved.
Any thoughts out there? Are Java and Flash the only way to do this?
Re:Use whatever works
cbrandtbuffalo on 2003-11-20T13:46:34
I know you're big into graphing information, so I'd be really interested to hear your suggestions. I know you're pretty busy right now, but maybe when you get back you'd be able to take a look.
Basically, they are trying to create an undirected graph with nodes and connections between them (lines or edges). Burke would like to see it in 3D, on the web, and somewhat dynamic. They've got some demos with the look, but not the true functionality.
At its base, this really seems like a problem from the math world, so I'm thinking someone has already come up with a solution that could be easily modified (maybe a graphical turing kit)?
I'll have to start looking around.Re:Use whatever works
brian_d_foy on 2003-11-20T17:43:29
I don't think they would be able to store the graph anywhere, unless they had a really small universe. The number of links will grow too fast.
Indeed, just storing all of the information itself takes up a lot of space.
Say we take any subject, like "Dogs". How many links to we show? Categorization becomes a nightmare. Do you just show breeds? What about personalities, like Lassie? How many ways can you group dogs together? How far do you have to drill down to get to the first link that you want to explore? This is a problem for someone with a Masters of Library Science, and they either work in libraries or at Andy Lester's company.
Besides that, I imagine a lot of the connections will be subjective, necessarily. A couple of months ago I read "A Modern History of Salt" (or something close to that). I found it curious that a lot of things that do not appear related (say, salt and the dominance of the British Navy) have unexpected connections, while things that look like they should be related (salt mining methods in various parts of the world) really aren't because people came to the same methods separately. I would have never thought to link something so common to so many other things.
Ultimately, I am completely disinterested in the venture because why should I believe anything the Knowledge Web says? I am sure that it would have all sorts of interesting things to say about this little war over here, and most of it would be sensational. Very little that I read or see about the events over here even get it right. The problem seems harder from a historical standpoint rather than a technical one.
If someone wanted to actually do this, I think they should worry more about where to find the people with the knowledge and the people to organize it than which flashy-whiz-bang-here-today-gone-tomorrow application wil display it.
Then, they will have to take into account that knowledge is not fact. Iraq and the US have very different ideas about why I am in Iraq, for instance. There is no one answer.
That said, Jon Orwant, at one of the first Perl Conferences, showed a Java Applet that displays multiple directory levels at once using size of type, coloration, and position on the canvas. I forget the name of it, although I have a hacked version somewhere. I just saw a similiar thesaurus application for MacOS X in whatever MacAddict issue they have in the library tent (the issue has boxing gloves).
The Knowledge Web sounds like a thesis projects, and thesis projects are thesis projects because nobody else wants to deal with those problems.:) Re:Use whatever works
inkdroid on 2003-11-21T15:11:23
The Knowledge Web sounds like a thesis projects, and thesis projects are thesis projects because nobody else wants to deal with those problems.:) I loved James Burke's show growing up. I was always amazed at how he made those connections. I doubt he started with two events/people etc that he wanted to connect, and then went about connecting them...which amounts to the travelling salesman problem.
From what I saw at the KnowledgeWeb site, it's going to be an exploratory system, where each topic is then linked to others, and the content is user contributed. Sounds like the WikiPedia to me. I wonder if James Burke would like the WikiPedia...
