[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [Elist Home]
Subject: RE: [humanmarkup-comment] modified HumanML taxonomy
For David's modified taxonomy to work within the current design, the classes need to identify the primary, then derive what he has now from that. Because we still have work to do to ensure the primaries are what is needed, anything is just a draft or prototype of an approach. That said, the prototype is neat. How to apply theories of behavior will be needed. The diagram I submitted yesterday (derived from a paper on "Evaluating Intelligence: A Computational Semiotics Perspective: Gudwin" which references a paper by Albus, "Outline for a Theory of Intelligence" is a rough sketch of the kinds of layers one might apply which can take a sign system as input and react intelligently given some environment in which the sign consumer acts and reacts. Whereas we may not specify a behavior system, per se, we do need some overall context of how we posit the language we are designing will work. The output of such a system IS a behavior, an action, and that might be one specified in a secondary taxonomy of behaviors. Such behavioral systems are used in many fields for prediction and analysis. A question for the current phase is how we propose to declare human communication behaviors at the most useful abstract level, that is, what goes into the primary set. It has to be a pretty simple abstraction. Remember everyone, this is out of the box, artificial. We can make models and discover by prediction and observation contexts in which they are useful and accurate, but to say we have created a "true" model of humans or even fully described human communication would be a big stretch at best. len From: Emmanuil Batsis (Manos) [mailto:mbatsis@humanmarkup.org] paul wrote: >And, as always, there are more than one valid theory of behavior. > IHMO, there is no reason to cover any theory of behavior. It's impossible to invent something that will answer to every need and interpretation; the best one can come up with, is a reusable set of tools to use for vertical development. >The issue of how RDF classes are expected to be used is also important to talk about, and all I ever see is conversation among the initiated as if some first principles where understood by everyone who needs to understand. > I agree with your view that we are all guilty on that, however this is volunteer work. If one is interested on some basic background, he must go out there and look for it. In a list with so diverse interests, providing the basics would be unbearable; the volume of messages gets awfully great and messy when compared to posts maching one's interest. >Now one can ask about the origin of this list of about attributes. Is the list complete in every way or are there some issues here and there with one or more of the attributes? > >Who has decided on these and not some other attributes? Can we add or subtract to this list? > How about "no list", just types from which you can draw your objects. The type, or maybe interface if you preffer, from which one can draw members of that type is the best there can be as far as my limited imagination has gone. Besides, no term can have a global, unique meaning amongst applications... or even people; thats easy to figure out once you ask for interpretations from some. People that give the same interpretations in a term, usually share a similar function or other relationship against it. So the only way for such a language to be usefull is to provide the tools; links for connecting types with handlers whithin an application to provide for emulation of meaning and anticipation.
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [Elist Home]
Powered by eList eXpress LLC