Annotea

[Kévin Ottens]

Hello,

* Disclaimer 1 *
I haven't looked at Annotea itself... From what I understand after quickly 
looking at the project page it's more or less a RDF dialect to annotate 
ressources.

* Disclaimer 2 *
I had a hard time recently to track the real state of Klink and where it's 
going, so in the following I might make wrong assumptions about it. Please 
correct me if it happens.

Le Samedi 19 Mars 2005 23:03, Scott Wheeler a écrit :
> From what I can tell here it would be bound by the same constraints as RDF
> itself.

I fully agree here... But it's based upon my experience of Annotea... so 
please refer to Disclaimer 1 and add more information as appropriate. ;-)

> I'm reading up on RDF at the moment (picked up Practical RDF from 
> O'Reilly earlier this week), but it seems that it's mostly inappropriate
> for KLink.  RDF is focused on conveying meaning rather than connections
> more generally (meaningful or not).

Indeed, RDF *is about meaning*... Don't forget that it's designed for semantic 
web. Semantic web people just want to annotate everything under the sun using 
pointers to ontologies concepts. And here enters Don Quichotte in my humble 
opinion.

I won't say this goal is impossible to reach or not... But I'm sure that 
currently for something trying to "draw" links between documents and/or 
meta-data at the desktop level, it's impossible to use. Simply because in the 
end you would require the user to annotate the document by hand most of the 
time.

> Kévin may want to jump in on this at some point as, well, he understand
> such things much better than I do.  ;-)

I'm not so confident about my skills. But, well, I'm trying to step up and say 
something not too dumb on this one. =)

> In a nutshell I think the fundamental difference between a KLink like
> structure and semantic webs is that KLink is more concerned with tracking
> unqualified relationships and whatever data that it can get a handle on

And that's a sane choice on the desktop level in my opinion. Finding 
unqualified relationships is "easy" (ok, maybe not *that* easy), but trying 
to qualify relationships is really at another level... it's really more 
difficult, and near impossible in an automatic way (except maybe for some 
particular relationships).

> along the way and that "meaning" is an emergent property of a web of
> relationships.

From what I understand of Klink (Disclaimer 2), I'd prefer to say that you use 
unqualified relationships... but manipulated by the user in a contextual 
fashion. There's no meaning in the system but the user will interpret them 
*himself* thanks to his current cognitive context. That's why we could say 
without being too wrong that the meaning of the "klink web" is emergent.

> RDF and friends are more concerned with providing a notation for describing
> resources.

Exactly.

> Basically it's a question of where the data becomes meaningful -- KLink
> says it's meaningful after it's out of the system,

That's why I consider that the "meaning" comes from the contextual 
understanding of the user.

> RDF says before it's 
> meaningful before it comes in.

And that's why it's so difficult... If you want the system to be really 
efficient you'd better not make mistake in your annotations.

> Or something.  (Note that this is just 
> after my experience of a couple days of reading.)
>
> One of the questions currently going through my head is to what extent one
> can be mapped onto the other, but I won't even try to answer that until I
> feel like I have a better handle on more of the semantic web structures.

I have really no idea on this topic... I'm not even sure to understand... You 
use qualified relationships or unqualified ones... mixing both looks weird to 
me.

Regards.
-- 
Kévin 'ervin' Ottens, http://ervin.ipsquad.net
"Ni le maître sans disciple, Ni le disciple sans maître,
Ne font reculer l'ignorance."