Re: ontology try and try again [was: Re: Viral fragment identity ecosystem]

On 10/4/05, Al Gilman <Alfred.S.Gilman@ieee.org> wrote:
>
> At 3:48 AM +0000 10/3/05, Steven Ellis wrote:
> >Hi,
> >
> >I think that document fragments, whether they be structurally or
> >conceptually coherent, need to be permitted formal expression in
> >XHTML. The simplest illustration of this may be xhtml microformats.
> >
> >Bracing microformats (and other fragments) using globally unique
> >identifiers would permit machine isolation and analysis, cross
> >referencing, consensus, scriptability, coalescence, and the
> >association of folksonomies with high resolution.
> >
> >Please consider an attribute capable of accepting an arbitrary
> >'unique concept / null concept' identifier.
> >
> ><div identity="3C05DC85-DC34-4546-9210-02EC43188367" id="MyCard"
> >class="hCard">Content</div>
> >
> >In this case 3C05DC85-DC34-4546-9210-02EC43188367 may achieve
> >consensus as an hCard microformat reference.  Can you speculate how
> >this will scale? I thought it good enough to share.
>
> You are right that @class has been coopted for something less than it
> was designed to be.
>
> The HTML WG is taking a second run at ontology with @role. Now at
> least we have QNames where there is a presumption that any value you
> give to @role has a public explanation somewhere, as opposed to the
> private-code @class tokens.

I see two divergent forces here: 1) the natural language that people
communicate in that is very much based in statistics and social
networks and 2) the signal loss that is required when computers get
involved. Computers, to date, have only been able to act on
information when information has been filtered down into something
common. Computers can only work on something when they understand it;
at least when they have some sort of explicit link to it.
Traditionally this has been done by a programmer or web designer
somewhere.

The trick is to find a way for interactions and behaviors to be
attached to data using social networking techniques as opposed to
categorization systems. This would mean, for example, that a computer
would have it's own interpretation of words and that those
interpretations wouldn't necessarily be consistent from location to
location. However I think that many people enjoy the fact that the
computer acts consistently from session to session and computer to
computer.

I think what we have here are two conflicting goals; one, the human
desire to be fully expressive; and two the human desire for the
computer to be able to do something with that expression.

And end goal as I see it would be this: don't limit vocabularies and
tagging, but do strongly type collections of information that the
computer can interact with, slowly expanding on them as computers
become more capable; e.g. don't limit what one can put in keywords;
don't limit general ontology. Do ensure that the computer clearly
describes what it can represent; e.g. contacts, documents, slide
shows, images, etc. Then force users to specify relationships between
objects (either implicitly or explicitly) between objects producing
computer-comprehensible relationships. I.e. the computer knows what
the document is, knows who the person is and knows that the document
was reviewed by the person.

And this is basically what some people and companies are already
doing. Let tagging reign supreme, but do create types of objects users
can interact with like contacts, documents, devices and so on.

Now the problem with role is that as I see it from the two camps: one
believes it will bring order and consistency to the world; the other
believes it will reduce my expressiveness. I believe both are right.
Which makes role for the sake of itself a bad idea. Role, however,
limited to the number of things that computers can commonly operate
on... that's a good thing. So let's limit the values of role to those
things we see and are likely to see in the future based on what we've
seen consistently in the past; e.g. footers, headers, navigation,
content, asides, etc. Then let's put each thing with a role into it's
own requestable stream and separate them out from the core content.
No? Well I thought I'd sneak that one in there.

--

Orion Adrian

Received on Wednesday, 5 October 2005 12:24:35 UTC