Back ... ... ... ... ... ... ... ... ... ... ... On Stream ... ... ... ... ... ... ... ... ... ... ... Forward

OntologyStream Inc.

The Gelhard-Prueitt Architecture

Nan Gelhard and Paul Prueitt

5/21/01

(draft)

Introduction

Artificial intelligence attempts to mirror some aspect of the world. These systems create a machine processing space that serves as a proxy for the world. Control of the proxy is translated into control of the world. This principle is thought to be true regardless of whether the aspect of the world is a simple pendulum system, a military battle, or an Internet commerce transaction.

In the current information technology paradigm, knowledge is said to exist in a space of machine addressable subjects formatted according to machine rules. The processing is limited to computational processes, not because the system cannot be open to human interaction; but by design.

In the Gelhard-Prueitt Architecture, we are not talking ONLY about a model or mirror world machine processing space. We talk about how a machine addressable space might have proper interactions with the processes that occur in the human mind. We hold that processes in the human mind are not computational or even computational like.

We are developing architecture to integrate machine processing and human processing. We look for good ways to make the machine assessable space work according to human norms instead of making humans access the machine by processing information according to machine rules. Clearly, our specific notion of a machine addressable space must have specific qualities. Knowing what these qualities are has been problematic for the industry.

We anticipate that fundamentally different computing ideas are developing. Our hope is that these ideas involve the encoding of knowledge about processes into XML structures. With the conceptual model of Topic Maps, we feel that the addressable space might have proper interactions with the processes that occur in the human mind.

Section 1:

Computational Knowledge Spaces are being developed using machine-readable ontology and an abstraction layer that controls the inner linkage (configuration) of these machine-readable ontologies. The abstraction layer is the process model. It allows one to stand above the topic map and to make an interpretation of the meaningfulness. Throughout history and within many disciplines one will find work on architecture with a second order control space. The second order system allows the control of the first order system. No higher order is needed because the second order system has human perception as an essential element.

The entire web itself can be seen as a (rather poorly developed) machine-readable ontology. What we are all looking for is some way to make this organization more acceptable and easier to work with.

A machine-readable ontology can be huge and without meaningful ways of interpretation. However, a projection (of a specific aggregative type discussed in Murray and Prueitt (2001)) can be made from any ontology – whether well formed or not. A formative projection is defined and constructed from the user's point of view. This notion is discussed in the URLs given below.

Some very basic concepts can be made to help us in the process of bringing meaningful routing and retrieval technologies into existence. These concepts are likely to be very simple and surprising. An example is given regarding the conversion of data into various number bases. One can convert all data in each field of a data source into base 96 so that the ASCII text elements are numbers (base 96). Treat these numbers as part of a number line, a tensor of rank 0, with the natural order. Occurrences of the number “John” will occupy the same position on this number line regardless of which data source the number “John” comes from.

Now scatter gather and encode emergent relationships into tensors of higher order.

There is more here, such as:

1/3 in base ten cannot be represented finitely as a decimal. However, if one converts this "quantity" to base 6 then one will find that 1/3 = 0.2. An infinite process is avoided. The base does not change the appearance, but it does change the arithmetic processes that occur. These judicial changes in arithmetic base means that round off error in computers can be easily avoided - if we only understood finite computing.

The quality of being prime, on the other hand, is invariant under number base conversations. The proof of this is quite enlightening, again with deep implications.

Generalizations of this base conversion theory can be applied to the issues related to language independent knowledge expression. These language independent expressions have to be projected into a scope that limits the interpretation into a pragmatic and semantic form. A categorical relationship is thus established between elementary number theory, natural language processing and general problem solving.

Section 3:

A formative projection of ontology is a theoretical notion at this point. However, we at OntologyStream Inc. believe that distributed web architecture for these projections are well specified.

Funding, not only for us but also for the entire paradigm, has not been forthcoming. The concepts are clear, but the concepts are not consistent with the mainstream IT paradigm (funded by NSF and DARPA).

As background to this discussion, four URLs are given:

http://www.bcngroup.org/area3/pprueitt/forms.htm

and

http://www.bcngroup.org/admin/CIL/guide/events.htm

are works on web architecture done a year to three years ago.

http://www.ontologystream.com/OS/MarketDelineation.htm

on In-Memory databases and structural holonomy.

http://www.ontologystream.com/OS/workFlow.htm

on work flow and topic maps.

In Summary

A formative projection focuses and limits the view of a machine addressable ontology to what a human can apprehend. The projection also places this view into a process to be expressed over time and in a conversation with the human.

The formative projection is required to have certain human like properties. The projection must arrive at an appropriate answer. In some cases, variation in viewpoint will require multiple answers – each of which are validated within the scope of the viewpoint. The separation of projected views of the machine-readable ontology is the most critical feature required of the Gelhard-Prueitt Architecture.

The two deepest technical issues that we must address are (1) the issue of consistency or explanatory coherence, and (2) the issue of completeness.

The ISO Standard for Topic Map is a girder structure framing data sets for processing. However, a processing framework is required for swappable sub-structure and formative projection specified in Murray & Prueitt (2001). Such a process model for Topic Maps must regard the set of all Topic Maps like a room with movable furniture, or a library with a changing collection of books. The process model must be an overlaying structure that enables the processing of Topic Map data sets while interacting with subjects. This second order system stands away from the topic maps and allows the human subject to make a judgment regarding the meaningfulness of the topic map.