I've been taking time on the weekend to check on the state of the "semantic web". The difficulty in getting content authors to agree upon and then use a common taxonomy of terms is very real. But I'm encouraged to see that there is good progress is getting machines to make semantic relationships for us.
Semant-O-Matic is an experimental latent semantic search engine that has indexed a collection of 1400 blog posts collected from a handful of bloggers between March and June of 2002. The results are impressive. The search engine effectively determines associations between words and provides matches on documents even when the requested keyword is not present -- the semantic index knows that certain words are related.
There's a good acedmeic piece, Patterns in Unstructured Data -- by good I mean it's readable :)
Google thinks this technology has legs, so its definitely worth a look.
The experiment in LSI has an opensource perl script I'm trying to get my hands on -- with any luck I can see if this thing works on Fullasagoog.
Posted by modius at 10:30 PM | Permalink
Trackback: http://blog.daemon.com.au/cgi-bin/dmblog/mt-tb.cgi/111
