Re: [kDev] KendraBase Progress update...

[Daniel Harris <daniel@xxxxxxxxxxxxx>]

On 24 Oct 2003, at 15:33, Daniel Harris wrote:
> Neil and I have been making some good progress. We're going to start  
> posting notes here:
>
> http://www.kendra.org.uk/wiki/wiki.pl? 
> KendraBase_Design_And_Implementation_Notes

Had a semantic revelation while sitting in a forest as the sun went  
down yesterday. I was having a problem with understanding how an HTML  
file would be stored in our triple store. Of course you can just store  
it as one piece of free text. But what happens when you want to relate  
one element in it (say a piece of display text) to a dynamic item in  
the database? On the Wiki I wrote:

Displayed as:       Stored as:
(ABCDEFGHI)   <->   (A)---(BC)---(D)---(EFG)---(HI)

But I was dividing up the text file *in a linear fashion* to  
semanticise the object (say 'D'). When Neil said it could be a tree it  
didn't sink in. But now I know! An HTML file is a semantic document  
already. When we put it into our triple store we don't need to do it  
just text wise (although we can if we want too) but we can rip it apart  
based on its meaning. Each element and attribute will have its own node  
(as in node---link---node).

That means we can take any XML document and do the same thing - fine  
grain map it into our store. The powerful thing about this is that we  
can get it back in different ways depending on our filters. Powerful  
stuff!

Also, taking this to the extreme - so not in our project scope, you  
should be able to take any script in any syntactic language and place  
it in the triple store. And, assuming that we've got equivalence  
statements, bring out the script in another syntactic language. Oh mad!  
;-)

Anyway, that's helped me a lot. I feel the need for some snazzy 3D  
diagrams to help us understand how all these many facets form  
KendraBase.

Cheers Daniel