Monday, September 03, 2007

3 September 2007: Cardiff

In Cardiff for a couple of days attending my last conference as a Research Fellow.

The last three months has seen me work, sometimes pretty frenetically, on a UIMA-based tool to take web pages, strip out the tags, investigate the text and populate an ontology (kind of like a database in this case) with information about terrorist incidents in Afghanistan, so that Dong-who-sits-opposite can use it to put dots on his map. Somewhat amazingly, it actually does seem to work, and so today we're showing a live end-to-end demo where, at one end, you stick a couple of words in a search engine (usually "Helmand" and "Incident") and it churns away for a minute or so, sending messages back and forth between servers located in various parts of England and Wales, and then boom, there it is, dots on a map showing the incidents reported by the documents you searched for. Neat, and not totally inaccurate either. And it's pretty much all programmed in Java, no Perl in sight, just a smear of PHP near the surface.

The fun thing is that, as our demo requires internet access, our colleagues based here at the University of Cardiff cunningly (and secretly) negotiated for our specific project to have a little web access so we can show our demo. The rumour is starting to spread that there's internet over here in this corner of the room... the looks we're getting from the other projects...

Course, it means I can blog a little while we wait for the presentations to finish...

4 comments:

Anonymous said...

Quick! use it to find the weapons of mass destruction!!

McDougal

Anonymous said...

actually reminded me of http://www.chicagocrime.org - hours of fun, better than genevacrime.org which is just a map...

Blaine Downie said...

sorry for my ignorance, but what is the practical purpose?? Certainly not meant to diminish your work Duncan!!

DuncMcRae said...

No, diminishing is fine with me.

Roughly the story is this: the customer here is, probably, somebody military and this tool is designed to be smart enough so that they don't have to have people spending loads of time wading through grey shoals of news stories, web pages, blogs etc to build up a picture of what's going on... the 'terrorist incident' area is just an example, you could do it for anything, pasty shops for instance. Although they don't make the news so much.

Of course, I don't have so much to do with it all these days...