Base Schema-Address and Re: [humanmarkup-comment] [humanmarkup TC]Assorted meeting notes - from Rob

[cognite@zianet.com]

Intro.  

Now the HumanML committee has moved [from "phase 0", discussion] into
the stage of actual schema design for HumanML. 

Thus saith Rob Nixon:

At 11:38 AM 19-05-2002 -0500, you wrote:
>Hello everyone, sorry for the delay.
>
>Here are a few notes related to the discussion we had during our monthly
>HumanML Technical Committee conference call on the 15th.  Many of these
...>speak in the
>languages of physics, mathematics, and systems sciences, 
...
> also overlaps with the concepts underlying the semantic web approach.
>
>These notes are not meant to send us wildly off course, but rather to
>make sure that we have explored our assumptions.
>
>1). ARTIFACTS:
>
... [marvelous exposition, q.v., referred to in part below]....

Per Rex Brooks and Ranjeeth Thunga's guidance, we've begun jointly
re-working through the caderie of possible terms for HumanML. Today in
our conferenced telephone meeting note was taken that since our problem
is handling the "human" which is invariably also contextual, a shallow
list of terms such as suitable for some other markup endeavors is
patently insufficient. How, then, to proceed to bring contextualization
into our base of primaries (cf. primitives)?

The following analysis of the first term under discussion, ADDRESS, sets
it up as RELATIONAL, with CONTEXT CONDITIONS. 

It turns out to be amazingly consonant with Rob Nixon's concurrent
discussion of the second term, ARTIFACT. 

Rob's point j) describes how an ARTIFACT is "manufactured" ... "out of
the knowledge and information field of the individual or the group". His
points g) and h), speaking of 'trajectories' zooming "through the
knowledge and experience 'space' of the individual perceiver" [who is,
themself,] "a node in a cultural and social network...with (feedback
loops) interconnecting the artifact nodes and beliefs) among the
interacting individuals" propose the interesting idea of "momentum"
within a dynamic net.

When an ADDRESS is a special case of an ARTIFACT, as I'd opined in our
phone meeting, Rob's description of ARTIFACTs in terms of nets are thus
illustrated by the concrete "semantic-net-portions" for ADDRESS set out
below.  Elaboration on contextualizable nets follows, and then some
questions to work from.


comments re humanML base schema term or "region" of meaning:
											c. by author:  S. Candelaria de Ram, 15 May 2002
--------------------
contents:
	0. Intro
		dialog, point of departure
		TOC
		bridge
	I. analysis for ADDRESS, working term currently under discussion
		working notes
			REF, sample current apps' "address records"

convention for expressing context in [processing] rule by using x ----/C ----> y

			generalities extracted

contact method ---/situational conditioning 
---> contact location specifier(s)

referror [signifior] ----/situation including current ROLES 
----> reference NAME(s) or DESCRIPTION(s) [sign/symbol]  [of REFERENT,
signifie']

			discussion points, A B C
				pan-culture
				agency of processes
				ADDRESS as special case of next term ARTIFACT
	II. general theory for implementation (of contextualized relational nets)
		approach background, dependency graphs
		REFs, hetnets, ghetnets [grounded] heterogeneous nets, v. semantic nets
	III.  Schema Design, development questions, A - H...
		(nature and interrelation of HumanML PRIMARIES and SECONDARIES;
		ancillaries; uniformity/content-direction issues)

--------------------

The following analysis is open for plenty of discussion.  The questions
in the last section particularly are directed toward large-scope issues
of design for making a COHERENT SCHEME.

================
I. analysis for ADDRESS, working term currently under discussion

working notes: 

	REFERENCE for comparisons: Dan Conolly's 
	www.w3.org/2000/10/swap/pim/contact.n3 
	which discusses MS vcard, MSOutlookContacts.n3, q.v.

	convention:  / for in-condition-of or CONTEXT specification, as 
		used in linguistic rules like this:
			PRONOUN.demonstrative --> these / near & plural
												this / near & singular
											  those / far & plural
											   that / far & singular
		These conditionals may be formulated alternatively, for instance
		as logic rules or equivalently as nodes/links in a reasoner net.
	

	generalities extracted:
	
		contact method ---/situational conditioning ---> contact location specifier(s)
		
			e.g., computer browser ---/connectivity etc. ...---> website address 
									as domain name or IP number and optional directory location
				Note implicit action from computers on both sides.
			e.g., postal mail ----/situational conditioning ----> postal address
				Note implicit actions from matrix culture and service agents on both sides;
				multiple postal addresses (office, home, vacation home) and contact
methods usable for
				different times, relations between communicants (interlocutors).
			e.g., shank's mares ----/situational conditioning ----> geophysical
access provisions and landmarks
				Note relevance in some conditions, agency involved both sides.
				
		referror [signifior] ----/situation including current ROLES ---->
reference NAME(s) or DESCRIPTION(s) [sign/symbol]  [of REFERENT, signifie']
		
			e.g., names such as first-name, ..., last-name
			e.g., role names such as Mom, Sir
			e.g., email alias
				
	discussion points:

A.	The generalities might be treated as PRIMARY/BASE/1o terms. They are
pretty much
pan-culture.  

The examples given (with "e.g.,") are specific to certain situations
of use or application.  Logically, that would push them to SECONDARY, 
application-specific -- or maybe [partially] SHARED-by-secondaries?
(Cf. the shared portions of gcc and other software suites.)


B.	Process is involved in utilizing ADDRESS; application for communication
therefore renders it neither solely noun nor verb. Whereas a node alone
is non-relational, THESE DESCRIPTIONS ARE RELATIONAL, AND
CONTEXT-CONDITIONED.  The "contact location" portion is the ADDRESS.
AGENCY is implicit in its functional significance.  An ADDRESS can't be
an address unless it is interpreted in real-world activity (NB 
logicians) as a location for contacting [somebody].  Process in this case
requires agency.  

Is it correct that mutuality enters in here in all cases, or not?  


C. 	related working term:  ARTIFACT

comment: A particular ADDRESS is an [instance of an] ARTIFACT. ADDRESS is a
kind of
ARTIFACT (cf. ISA, AKO, subclass). In having to do with significant
(maybe even purposive) activity it differs from being a plain location.
It has co-operative significance: That "human" social property ;).

================
II. general theory for implementation

A. As noted above, this analysis of sets a term such as ADDRESS
up as RELATIONAL, with CONTEXT CONDITIONS. 

That seems to be what we need. Whether this is necessarily
Object-Oriented is not clear, but it certainly would seem to be
codifiable using contextual reasoning representable as traversal of
enhanced nets or dependency graphs. ("Net traversal" is a way of looking
at proof sequences, though the proofs need not be in classical logic,
obviously.)

B. -- No doubt this approach is being influenced by prior work on things
like the "hetnets" that were developed just for such problems. (Hetnets
have heterogeneous links; grounded hetnet Systems are "ghetnets".)
Hetnets, for example, have been used for [reasoners] reasoning from a
mixture of knowledge and beliefs and thus employing typed links such as
strict/defeasible. 

[The most accessible reference is probably this one on hetnets:

Ballim, Afzal, Sylvia Candelaria de Ram, and Dan Fass. 1990. Reasoning
using Inheritance from a Mixture of Knowledge and Beliefs. 
Lecture Notes in Computer Science series No. 444 (Lecture Notes in
Artificial Intelligence sub-series). Springer-Verlag, Berlin & NY.]

The most relevant is in process of conversion to HTML from troff -ms,
which is turning out to be non-trivial.  But in the meantime I've 
recovered a Postscript version that can be posted at
cognitionandcommunication.cognizor.com which for some hours you
have to get to thru:
 
http://cognizor.com/CogYCom/index.html

Candelaria de Ram, Sylvia. 1990. Belief/Knowledge Dependency Graphs with
Sensory Groundings. Proceedings of the Third International Symposium on
Artificial Intelligence Applications of Engineering Design and
Manufacturing in Industrialized and Developing Countries. Instituto
Tecnol6gico de Estudios Superiores de Monterrey, Mexico.

NOTE FIGURE 3 in the latter article, which depicts how a complex of 
support steps can be viewed as a whole.  (It is a node incorporating
both a conclusion and its proof steps. The proof depends on the 
combination rules that link the current state to antecedent values.)
This shows how a single HumanML term, with its internal structure, 
can be context-bound.  

A term, seen as a ghetnet node, is sort of like a deep cross-section
taken at some point of perspective in a net. The types of things used to
derive a current value (the affecting conditions of the designated time
at the time of evaluation -- a process) are comprehended within the
term. Their inter-relations are expressed by "rules" or "constraints" in
a reasoner.

In analogy, a physics equation describes certain inter-relations among
parameters, where those parameters can have ranges of values. (A
constant has a fixed value; temperature has a starting point and more or
less continuous values; etc.) The parameters may get evaluated at some
point on some dimension(s). ( -- We want to do that too, to mark "human"
states.) In particular situations, there may be values or families of
values on one or more parameters -- like a ground-voltage on a metal
cylinder, or for humanML, like a social group's frame of reference:
"Boundary conditions" is the well-worn descriptor for these in physics. 
By the way, in treating the range of humanML at this point, parameters 
cannot be presumed to be orthogonal (independent of each other).

If the structure of a ghetnet changes, it's different than if the values
of the conditions do. For instance, the concept of ADDRESS has recently
expanded to include email addresses. But if the system rules change
so different kinds of ghetnets will be generated, this is a SYSTEM META-LEVEL
change.

For real life, the time of evaluation is NOW, whenever that is.  We
can re-analyze situations with past (or future) DESIGNATED dates ok, 
but are constrained to current evaluation.  

In Rob's example where a meta-processor in 2002 is figuring out a book
accession number for a book in pseudo-1930, even if the clock is wrong,
the time of evaluation is still NOW, isn't it? So the accession number
dates from NOW and is not a 1930 date, but rather an attributed
meta-processor-provenience-PROPERTY with a [dual] timestamp. Sort of 
like our need to tell kids what is imaginary and what's not. In a 
simulation world, plain dates and subsequently-attributed dates are 
still plain attributed-properties. The
critical distinction would seem to be that there is REAL time and
attributed time; REAL time is GROUNDED to -- real time.  We're not
making it up.  It runs on its own.  

Rob's idea of "momentum" is not developed in the ghetnet work I did
earlier, and "might could oughta" be a next step.... It appears to be a
very powerful concept.  

So the "prior cultural experience" has its effect, I guess!
================
III.  Schema Design, development questions:

A.  Is this contextualized relation kind of structure indeed what we need
for humanML?

B.  Will other terms have similar contextualized relation form?

C.  Is this the proper level for PRIMARIES?  Are we trying to make them
pan-culture ("culture-free", universal)?  If so, does this do it?

D.  Does anybody have any example(s) that would tell what level(s) might
be appropriate for SECONDARIES?  Didn't we say sometime that they would
differ according to their application purposes and might have
derivatives themselves?

E.  What about this idea of a SHARED pool of [optional] parameters suitable 
for purposes like transitions that take up PRIMARIES for use in [more specific]
SECONDARIES or ...?  In what ways do these differ from the GLOBAL
attributes mentioned in today's discussion, and/or are they the same?

F.  Will SECONDARIES use the PRIMARIES' [contextualized relation] form?
Will all SECONDARIES use the it?  

G.  Are there other forms that will also apply?  Or are TERM DEFINITIONS
uniformly of the same form?  Is this affected by their implementation?
Should form(s) be according to HumanML "rule"?  Per demands of
[SECONDARY] domain?

H.  hundreds more probably...but that's ok, as the mission of humanML
is directed toward elaboration!

SC

Thus saith Rob:

At 11:38 AM 19-05-2002 -0500, you wrote:
>Hello everyone, sorry for the delay.
>
>Here are a few notes related to the discussion we had during our monthly
>HumanML Technical Committee conference call on the 15th.  Many of these
>are my thoughts on the subjects discussed, so any “naïve” thoughts on the
>topics at hand are entirely my responsibility.   I tend to speak in the
>languages of physics, mathematics, and systems sciences, so I’m sure that
>there are other more appropriate ways to present this material.  There
>are also overlaps with the concepts underlying the semantic web approach.
>
>These notes are not meant to send us wildly off course, but rather to
>make sure that we have explored our assumptions.
>
>1). ARTIFACTS:
>
>a) The “meaning” assigned to an artifact can change over time.
>
>b) The derived meaning at any given time is associated with the cultural
>framework in which it is considered.
>
>c) There can be many parallel (in time) meanings assigned to an artifact,
>with each meaning deriving from different cultural (or group) frameworks.
>
>d) It’s possible that an Artifact can act as more than a noun in that an
>Artifact can act (and I would argue almost always act) as a “signal”
>within the perceptual field of the perceiver.
>
>e) As an overly simplistic model, Artifacts can be thought of as the
>nodes of a network, with beliefs acting as the connections between the
>nodes.  Clusters of these nodes and connections, can be thought of as
>context, with the entire network viewed as the knowledge and experience
>of the individual perceiver.
>
>f) By treating each network as a surface(of arbitrary dimension) we can
>add time into the model as a series of  stacked surfaces with the
>“artifact” nodes connected to their corresponding nodes in the surface
>“beneath”.  The evolution of the meaning of the “Artifacts” over time can
>be viewed as a series of vectors, where these vectors may fork, continue
>through, or dead end ( as the artifacts may separate into multiple
>artifacts upon examination, remain consistent, or actually be lost in the
>physical or in memory).  This process can be viewed as a type of Cellular
>Automata (CA).
>
>g) These connected series of vectors can be thought of as a trajectory
>through the knowledge and experience “space” of the individual
>perceiver.  You will also find that there is a type of “momentum”
>associated with these trajectories as groups of related “artifacts” and
>the connecting beliefs about those artifacts reinforce each other.   It
>takes more to shift the perspectives (in relation to the artifacts) as
>time goes on if they have been reinforced.
>
>h) It should also be understood that each individual perceiver can be
>viewed as a node in a cultural and social network (which is hierarchical
>in nature)  with (feedback loops) interconnecting the artifact nodes (
>and beliefs ) among the interacting individuals.
>
>i) Artifacts can also act as a pointer to a series of Metaphors, or in
>and of itself act as a “Metaphoric” node.
>
>j) In essence a (manufactured) Artifact can also be viewed as the
>“condensation” of “meaning” out of the knowledge and information field of
>the individual or the group.
>
>k) It is also important to understand that when we are dealing with
>“Artifacts” (objects) within Virtual Simulations, the concept of linear
>time and cause and effect can no longer be viewed as it has been
>traditionally.
>
>If for example I am running a series of simultaneous “Simulations” each
>based on a specific time period ( i.e.  1920, 1930, 1940, 1970, 1993,
>2002) and I share an (Artifact – a book, a building, a coin) “object”
>among them (that contains “Static Data Members”, “Static Member
>Functions” ) I will run into a problem with potential cause and effect if
>we use a simple linear view of time.
>
>The following example should highlight the problem:
>
>If for instance my six simulations utilize a class of object called
>“Book”, each of the six simulations will contain their own object
>“instantiations” of the book class.  You can think of the “Book Class” as
>the Archetype of a Book, and each instantiation of the Book Archetype in
>each simulation as the “physical manifestation” of the Book Archetype.
>In this sense each of the books in the six different simulated periods
>have no relation to each other (other than “Bookness”) and therefore can
>not effect each other.  However, if we include data and functions called
>“Static Data Members” or “Static Member Functions” in our Book Class (
>Archetype ), then we create a link between ALL instantiations of books in
>ALL simulations.
>
>The reason for this is that the Static Data Members and Functions are
>associated with the CLASS and not the individual book objects in each
>simulation.  So if we had (for what ever reason) static data members
>called “Highest Catalogue Number” and “Date Assigned” which were used to
>assign the next instantiated books catalogue number in any given
>simulation,  all books everywhere in all simulations would access that
>“Highest Catalogue Number”.   Here is the problem,  let us say for the
>sake of argument that when we start our six simultaneous simulations (
>ie. Boston – 1920,1930, 1940, 1970, 1993, and 2002 ) that it just so
>happens that the first “book” object is instantiated in the 1970
>simulation.  The catalogue number “1” is assigned to that book instance,
>and the date of “April 5, 1970” is recorded in the Static Data member
>called “Date Assigned”.
>
>Now it just so happens that since the start of our six simulations the
>next instantiation of a book occurs in the 1930 simulation.  The local
>simulation sees that there has already been one book assigned, and so it
>updates the “Highest Catalogue Number”  to 2.  What it discovers however
>is that from it’s (the particular simulations perspective) the first book
>was assigned 40 years in it’s future, so in effect, it has experienced
>and effect from the future.  A simple time stamping of events in this
>case would lead to chaos and confusion.  Now if we update the Date
>Assigned for this second instantiated book to Feb 23, 1930, from the
>perspective of the 1970 simulation it has just had it’s past changed by
>something occurring in the 1930 simulation.
>
>This again is only meant as a simple example of my point.  The goal hear
>is not to pick apart the example or to say that no one would ever do
>this, or that this would simply be a bug, or to justify that these
>effects as being in entirely different times lines.  I am trying to point
>out that there can be non-linear, a-temporal effects in simulations and
>we must at least consider this as we discuss “artifacts” and “knowledge”,
>and “meaning”.
>
>The concept of time in this venue (and I would argue our own) can only be
>viewed as a series of events and not as a single linear sequence we tend
>to think of it as.  It would also be possible to set up a series of
>complex feedback loops that would involve interactions between the 1930
>and 1970 simulations that would be hard if not impossible to explain from
>the perspective of VR characters in each of those simulations.
>
>From the perspective of the VR characters, knowledge from the future
>would be mysterious and unexplainable.  And from the perspective of the
>VR Quantum Theorist, experiencing the bizarre effect of having  the
>results of a previously carried out experiment apparently fall into line
>with information only more recently taken into count suddenly becomes
>understandable.
>
>If our VR simulations are going to model our own weird “experience” they
>must incorporate mechanisms of this nature, and therefor require us to at
>least explore these concepts as we define a useful XML HumanML dialect.
>
>The previous points have been greatly simplified for clarity ( I hope ).
>The goal of the previous points have been to illustrate that the concept
>of an “Artifact” as a simple noun is insufficient.   I believe that
>rather than viewing an (artifact)/“Signal” as an interruption in a static
>field (as was discussed during the meeting), that they should be viewed
>as semi-recurrent / semi-stable dynamic “processes” (or eddies) in a
>fluid field (where “fluid” describes a dynamic network structure.)
>
>Regarding:
>
>2. ADDRESSES ( as well as many other attributes )
>
>We must allow for multiple concurrent addresses, as well as a historical
>list or tree of address ( again as we move forward and backward ) in time
>related to VR simulations (leaving out our non-linear time effects
>previously discussed).
>
>Again, these are all only points to consider.
>
>Rob
>
>
>
>
>----------------------------------------------------------------
>To subscribe or unsubscribe from this elist use the subscription
>manager: <http://lists.oasis-open.org/ob/adm.pl>
>
>