[rdfweb-dev] Re: Parsing FOAF in perl?
Dan Brickley
danbri at w...
Wed Dec 11 12:24:25 UTC 2002
* Danny Ayers <danny666 at v...> [2002-12-11 11:50+0100]
>
> >Its a sort of experience based reasoner. Like a human it will make
> >mistakes while it is learning who to trust.
>
> I've thought for a while that the logical semweb will benefit from (need!)
> tools from old-fashioned AI, though I hadn't really thought much beyond
> scraping. But the fitness function you describe can probably be lifted
> straight out of a neural net textbook.
> I guess until we have big flows of statements going on it won't be possible
> to properly test, but the idea sounds absolutely feasible.
>
So my 1998-era homepage was just
http://rdfweb.org/people/danbri/index4-nov98.html ie FOAF without the namespace.But the background graphic was more interesting; I generated it by
feeding a Kohonen Self Organizing Map (SOM) with some internet classification
data (from http://www.sosig.ac.uk/ I think). The SOM tools are out there
somewhere still, I believe freely usable, and work on simple text file
input. Unlike many Neural Net methods they don't require a training set;
SOMs just help you identify clusters and associations buried in complex
datasets. They've been used a bunch for Usenet mining, for eg.
See http://websom.hut.fi/websom/
http://www.cis.hut.fi/research/som-research/som.shtml
Hmm i can't find the software, hope its still up there somewhere. I'd
love to have an excuse to revisit this stuff (mutters about lapsed phd) but
I expect realistically I won't. FOAF-harvested data should be an interesting
thing to feed to such algorithms though, although the data isn't currently
that rich. If there were more Wordnet and Dmoz classifications being used,
for eg., I'd expect some interesting testbeds...
Dan
More information about the foaf-dev
mailing list