[huml] RE: LMNL toolkit for HumanML

["Bullard, Claude L (Len)" <clbullar@ingr.com>]

Really glad to see you here, Gavin.

It would be interesting/maybe popular to combine 
HumanML-derived codelists with a blog editor. 
People want a blog to be expressive.  If a blog 
were tagged for intent, emotion, etc., then it 
can be fed to a HumanML enabled avatar and 
hopefully someday, to a HumanML enabled avatar 
with a voice.  There are possibly more important 
applications, but my experience is that something 
that is fun combined with something that is 
popular is a good bet for viral applications.

The primary base schema is mostly XML Schema 
abstract types.  That is ok for loose categories, 
but a derived set is needed to do anything 
tangible.  I made them abstract because when 
doing the research into human communication, 
one finds mostly theories.  The categories 
are usually moreorless the same, but the 
theoretical intepretation (eg, the effect 
of proximity on communicators as regards 
dominance) differs.  So, for each abstract 
type, there could be multiple code lists 
among other kinds of markup.  A simple set 
which we got earlier with input from John 
Cowan is a set of emotions/attitudes with 
range attributes.  Emotions are somewhat 
like polarities; not solved, but managed. 
Given that, one can markup the emotion 
and then the rendering just does whatever 
it does given some intensity value.  That 
is a simple way to start.

len

-----Original Message-----
From: Gavin Thomas Nicol [mailto:gtn@rbii.com]


On Saturday 15 February 2003 02:27 am, Ranjeeth Kumar Thunga wrote:
> I really think that by allowing for multiple overlapping layers and ranges,
> we can seemlessly account for multiple interpretations and perspectives,
> embedded within the document.  Gavin mentioned that he would be interested
> in perhaps helping develop an LMNL toolkit for HumanML, and he could
> present it to us in some form (Gavin:  I'll let you take it from here ;))

Hi, and I apologise for getting back so late on this thread... things have
been rather hectic, as always. Anyway, from my (limited) understanding of
HumanML, I think LMNL is a useful syntax alternative to XML.

Overlapping ranges, occur in human communication, and more importantly, the
*interpretation* of communication, far more than most people realize. This is
a direct effect of the multiple interpretations of a given sequence of text
that is common: we all have a (however slightly) different way of looking at
things, and this is often reflected in markup, and the markup wars that
people have (it should be called a "p", not a "para"!).

Interestingly enough, given it's document-centric background, XML does *not*
handle overlaps and multiple interpretations particularly well (SGML was
somewhat better, especially if you used CONCUR and/or SHORTREF combined with
multiple SGML declarations). You *can* mark things up, with milestones and
the like, but the markup tends to feel somewhat unnatural. IMHO, LMNL should
be easier to author... but that's an opinion that needs to be tested in the
waters of real world application (FWIW. I think it's be easy enough to use
the Java Swing text component to write a LMNL editor).

I was thinking that it's fairly easy to do a LMNL->XML conversion that would
use things like milestones and/or attributes to tie overlaps together, so
that in the worste case, LMNL can ease the authoring side. In addition, I am
planning to build a formatter that should be applicable to LMNL. The
formatter would be able to produce either PDF, or somewhat ugly HTML. When I
mentioned a LMNL toolkit, this is what I was imagining.

I guess I should sign up for the mailing lists?

----------------------------------------------------------------
To subscribe or unsubscribe from this elist use the subscription
manager: <http://lists.oasis-open.org/ob/adm.pl>