Slashdot Mirror
University Launches Semantic Web Interface
kv9 writes "The University of Southampton has launched a new semantic web interface, called mSpace, that it says will make searching for information online, and learning about a subject, much easier. mSpace is a framework that gathers information sources and presents them to the user in a single window. It can potentially be applied to any subject, provided the basic information is available. The researchers say this means users will no longer have to wade through lists of undifferentiated data when researching a subject."
From TFA:
Imagine more than Google
Imagine a better iTunes
Imagine Google on iTunes
Perhaps my early brain development was flawed, because I'm at a total loss to imagine what a "Google on iTunes" would be like or even what that means.
I'm a big tall mofo.
...interestingly, the demo won't run on IE (at least, the versions I've tried, being IE 6 on default settings). Perhaps this is a sign of things to come - more and more applications just not running on IE, and preferring FireFox / Mozilla?
A stack or queue of filters, with select box GUI, and text+gfx output at the end. (oh, and potentially sound clips - edgy!)
The only "innovation" I can see is that you can add + remove individual filters. Which is not, so to speak, going to launch rockets...
I recal looking at a system (in java) that allowed overlay of viewports (little square windows) onto a graphic to add + remove filters (in the photoshop sense in this case). You could drag around these viewports and overlay them to get a venn-diagram like effect with filters (real time, over the web in an applet)
That was while I was a University (so was between 1993 - 1998, probably 96 at a guess). That was simultaneously; similar in concept, more impressive by far and much more of an "innovation" at that time...
I may be missing something, but I couldn't see anything "new" there.
Now that's a first. The classical demo requires a Mozilla-type (they say Gecko type) browser for the enhanced javascript capability. IE 6 won't even run it correctly.
While there are several approaches to accountability already present in the field, such as counting links to the data (like Google does), or having smart, professional, attractive moderators (like Slashdot does [1]), none of them are perfect yet, and I believe this is a problem that must be solved before any semantic web becomes useful.
[1] Ok, I may be karma whoring just a little here... ;-)
Meatball Wiki page on GopherProtocol
A copy of the Gopher FAQ
MacOrchard page with TurboGopher VR
I don't read ACs: If a post isn't worth so much as a nom de plume to its author then I wont bother either.
Although this project isn't strictly "wrapped around" (pardon the pun) Berners-Lee's semantic web, but rather an external semantic "space" defined by a conceptual foundation and then refined by users in the inteface, it still fails to address to metatag/metdata problem, namely:
1) The metadata sink. Creating an "mSpace" around classical composers is one thing. Doing the same on "quantum mechanics and philosophy" is another. As you broaden the concept, you have to depend on a more-refined framework of contextual and categorical distinctions. Eventually, you may be creating more metadata than data.
2) The metadata reflection problem. Metadata, in that it is not the data itself, cannot possibly reflect every notion, category of thought or context -- many of those things depend on the user's own interaction with the data (e.g. what you find "funny" I may find "dumb."). And, as often mentioned, metadata may in fact be missing, ouright misleading or incomplete.
IMHO, though metadata projects such as these are intriguing, the true "holy grail" of classifying data is understanding context. Thus, why worry about metadata when you have the data write there in front of you? Even a statistical anaysis of word/phrase frequency over say, 100 pages returned by Google on "quantum mechanics and philosophy" can yield concepts and connections without any metadata creation/foundation at all (i.e. the user analyzing the key words/phrases can make those connections on his/her own).
Clearly I'm biased, as I work on software for OS X that does just this, but still, I honestly believe that creating more data, just to describe what is an increasingly massive corpus (the web), is the wrong solution to the "understanding" problem.
Semantic web usually refers to back end semantics, that computers can use to "make meaning" rather than the kind you descibe. That's how the data gets associated in the mSpace to begin with. As for hierarchies, that's the final step of the process: the demo music space is a multidimensional space. How do you represent that space effectively in a browser? there's good research to show that we can handle 2d hierarchical representations well. So, the mspace slice lets you take a projection through an ndimensional space, flatten it, and get a temproary hiearchy. As for bach, you don't need to know about bach being a classisist - you would need to know he's a composer, but we could fix that with a keyword search. Move composer to the first column position, and select bach. when we have more associated data from the british library's news archive, you'll be able to go through reviews and historical accounts to get at historical perceptions of bach. in the interim did you know that glen gould's interpretation of bach launched a come back of bach to the canon in the 50's? thanks for taking the time to think about the project. I agree where we want to get to is an easy exploration of meaningful content.
When Bach and Handel show up in the Classical Era (they were Baroque) and the interface looks like it could just be showing results of traditional SQL queries, then naturally our attention is focused towards the manual tagging of information and so it looks like the project is showing us nothing new: "See, someone mis-tagged Bach and he is showing up in the wrong list box
All sigs should be as funny as possible, but no funnier.
Part of the project is to allow wiki-like connections to the info views for publishing related content. Or using talkback like pings to talk with brokers/aggregators so that mspaces can be generated dynamically, and fed dynamically based on available rdf.
This is very much a start - a look at what might be/come something (more) useful, not as the done deal
That said, we hope that for the interim, by having a dump out to google on a topic you've already identified of interest that that will let you explore more readily or associatively.