Slashdot Mirror
Welkin: A General-Purpose RDF Browser
Stefano Mazzocchi writes "Many consider the Semantic Web to be vaporware and others believe it's the next big thing. No matter where you stand, a question always pops up: Where is the RDF browser?
The SIMILE Project, a joint project between W3C, MIT and HP to implement semantic interoperability of metadata in digital libraries, released today the first beta release of a general purpose graphic and interactive RDF browser named Welkin (see a screenshot), targetted to those who need to get a mental model of any RDF dataset, from a single RSS 1.0 news feed to a collection of digital data."
The question is about whether we really need a World Wide Web that looks like Wikipedia with links to every word and generally just a jumbled mess of blue and purple text. No matter how you cut it, the problem lies in having too much information immediately available.
Imagine you are a reading a book, but each word is connected by string to a dictionary reference, and each dictionary reference definition is tied to the definitions of the words in the definition. You'd end up with a huge, eventually circular mess of string and you couldn't realistically get any enjoyment out of the book. The fact of the matter is that if you want to get more information about something, it is easy to go to an outside source to look it up. It does not need to be easier, because by making it easier than it must be you necessarily end up cluttering the thing you want to illuminate.
There is an old saw, "Make things as simple as possible, but no simpler." The Semantic Web, while an interesting idea, tries to make things too easy, beyond the point of usefulness. The lack of content on the Semantic Web is a testament to the uselessness of such an over-engineered web space.
Isn't RDF much like the laser use to be? A solution looking for a problem?
Or a testament to people's inability to understand new paradigms.
Or a testament of the inability of the paradigm's creator to get people to understand it's necessity.
"Even programmers aren't used to the idea of describing knowledge - it's one thing to tell a computer what to do, it's another thing to tell a computer how to know about something that you know."
Yea. Just try getting a programmer to explain the latest thing they're working on. "Well you see it does this, and if you click on that, something happens. It's all too complicated to explain, sorry."
Much work on the semantic web has been with n3
N3 is a superset of rdf, allowing for quoting of groups of triples (known a formulae). In n3, you can say things about groups of n3 triples, including about their trustworthiness.
For instance, you can say:essentially saying that the formula which is the semantics of the given document if of a class
There are many who are very wary on n3 for precisely the same reasons.
Note that I will always plug n3, given that I'm heavily involved with cwm.
Check out Semaview Inc. who's making a business of RDF. They've already got one good product out. They're somewhat OSS friendly, too.
Personally, I think eventSherpa is pretty neat.
(Disclaimer: I know the CEO.)
-kidlinux.
One trouble regarding many semantic visualization techniques involving large datasets is: the more visually appealing a graph is rendered, the less useful it often becomes. Many projects undertaken over the past 6 years (including Welkin) have focused on 2- and 3-dimensional renderings of a dataspace, using lines, proximity, node-shape, fly-over metadata display, etc. to classify and relate nodes, only to find there is no room left for persistent display of the textual metadata that ultimately drives a user toward the content he/she is looking for.
Marcos Weskamp's Newsmap (slashdot) on the other hand demonstrates an excellent balance of form and function, emphasizing textual metadata over symbolic graphic representation. How might this approach be applied specifically to RDF? One possibility: 5 axes rendered in a 2d visual space: color (category), saturation (relevance), size (interest), x/y position (age) and text (metadata). Just a thought anyway.
Yeah, why is this funny? I wasn't trying to be funny - this is a valid question.
All this talk about GBrowser (Google's browser), for instance.... what do you think it is going to be based on? Firefox, of course! You don't think Google would be crazy not to make use of that powerful, flexible, extensible platform that runs on all major operating systems, has support of geeks and hackers to the point where they contribute $250K for The New York Times advert.
Simpy
Statistical text analysis and link analysis are a superior technique because it presumes the author could be BSing. The entire document must contribute to the corresponding query value, not just keyphrases which could or could not be true. This is why Google is a $50 billion company and no RDF firm ever will be.
Well, you would certainly know, ease of use is critical to winning developer mindshare and promoting adoption of technologies - and I would point to SAX as a great example of this for promoting use of XML in the Java community.
It does seem to me that the key thing is to promote ad-hoc use of a relatively standardized mechanism for relating XML document structures to other XML document structures. Forget about waiting for somebody else to build relevant ontologies, reconstructing the entirety of human knowledge from the ground up, or any of that stuff. What people could reasonably do today is relate XML schema one to XML schema two because they need to connect widget A with widget B. Make the adoption of this technology as low cost as possible.
Just like adding a few anchor tags to a basic HTML document is an easy way to relate some human readable information to other human readable information, relating XML document types to other XML document types should be "easy".
Then the only big problem is to find a few applications that would actually demonstrate the benefits of doing this clearly. Yes, it is effectively a distributed XML database of sorts, but what is it good for? RSS has real applications for end users, so it has caught on. Without some software to demonstrate the benefits of linking up your XML data structures, people just won't bother with it. It seems specific, realistic use cases are what's needed here (and what seems terribly lacking from all the W3C RDF documentation as well). How does the distributed, semi-structured database that results provide use to me beyond what I have now with lots of disparate XML documents out there, when you cut out the truly grandiose notions behind RDF and the full-fledged semantic web?
I'm too tired to come up with convincing arguments right now, so hopefully somebody else will fill in the blanks here.