Post-meeting with Johanna

Had an excellent meeting with Johanna. In essence, she saw a way around the main bottleneck I'm having with the corpus (our rather erratic annotators) by simply sampling the reliable annotators.
She also clarified that we should have a "mid-level" semantic representation between propositional and full-scale predicate calculus for our machine learners. Now if I can just get SVMTorch to work :)

Argh!

Too much to do right now....currently trying to get this corpus of children's stories working, and while there definitely seems to be progress, I'm having issues with SVMTorch and LSA. And I'm not really sure how to do statistics for this data. Hopefully Johanna will have some ideas.

Young Service-Oriented Researchers Conference

I gave a paper yesterday in Leicester at YR-SOC 2005, given the somewhat dramatic title of "The Ties that Bind: XML, the Semantic Web, and Web Services"

Here's the slides for those that are truly interested.

Note that I do not use preventing in my slides...even my slides are REST friendly and Web compliant.

Yours truly at YR-SOC 2005

My fundamental thesis is that the Web is moving from a "universal information space" to a "universal computation space". To be a universal computation space, you need a universal (i.e. commonly accepted and standardized) way of doing things that one needs to do to actually compute. To compute, one needs state (and a serialization encoding for that state) and functions to transform state, and types for the state . One also might want to distinguish between the semantics of the state that are internal to the computer and those that are about the external world beyond the computer.

What I believe is that XML fundamentally provides the state, Web Services are functions (or at least should appear to be so to end-users!), and the Semantic Web can be used to model the world (ontological typing) while XML Schema can be used to type the state (syntactic typing).

Me speaking again, looking rather bedraggled

The talk got people quite excited, and I had such witty one-liners as "Reasoning over just XML is like reasoning over ASCII!" and "Let's face it, no one is using the Semantic Web because there's no data out there in RDF, so why not just use XML." I seemed to convince about everyone, and had a great lunch with Stefan Decker. Stefan explained to me some key differences between the original RDF vision (loose data connections) and OWL (strict knowledge modeling, doesn't scale). He also agreed that my vision was good, even if he would prefer people to just exchange straight RDF. Lastly, he made some encouraging comments about his idea of semantic desktop - which is exactly what I want to do for my Ph.D. thesis, show that the Web should be used to organize the locally rich semantic space of everyday users.

Links Meeting in Edinburgh

The Links meeting in Edinburgh was very interesting indeed. Somehow Phil Wadler had got a few dozen of the world's brightest programming language theorists (including Benjamin Pierce and Xavier Leroy) in the same room. The goal was simple: to create the next big programming language: Links. The presentations were excellent, but it felt like the people had yet to even be sold on the concept, yet prepared to co-ordinate working on it. However, if anyone could create the next programming language, these people could.

Who's that man in black lurking in the background?

Phil's goal is admirable: we need a programming language that takes the traditional three tiers
of Web development (database->script->XHTML) and replace them with a single unifed programming language based on the latest and greatest results in functional programming theory. Links seems at first glance like HaskellPHP, but Phil has greater plans. I think overall the vision is great, but the web development crowd already uses scripting languages, so we'd have to sell them on something else. I think the magic answer is Web Services.

The cake for Links

Phil cuts the cake.

Web Services are nothing but functions on the Web, and obviously a functional programming language would be better than composing them. The alternatives seem to be BPEL, which is huge, has no formal semantics, and is not even coherently implemented, and OWL-S, which is rather academic and also seems not to be catching on. Because to make Links work we need not only the academic community, but the hacker community to jump on board. And the hacker community is currently just toying with Web Services, and upset at the standards bloat. A simple, clean, and functional solution would sell it, and do something PHP can't do!

My Links links are here:
http://www.ibiblio.org/hhalpin/homepage/notes/links.html

And here's my original comments on Links on Phil's blog: LISP for the Web!
http://wadler.blogspot.com/2005/04/links-meeting-at-etaps-enter-comments.html

In other news, I volunteered at ETAPS, heard some talks, and got a free hat.

Me in my free hat...would you trust this man with your conference?

First post

I would prefer to host this on my own server, but accidently created this in a fit of late night blog posting, so will keep it till I get my own server. Testing, testing....anyone out there?