Friday, November 19, 2004

Three Types of Ontology for Crisis Management

Physical infrastructure

1) physical resources related to and used by processes that monitor, or that will monitor, shipping containers.

2) physical resources related to and used by processes that make responses to crisis involving the material contents in, or likely to be in, shipping containers.

Real time conceptual mapping

3) the real time human information traffic arising from normal operation of monitoring and response mechanisms

4) the real time human information traffic arising from operation of monitoring and response mechanisms during critical incident response.

Work-flow or process mapping

5) repository of normal shipping behaviors

6) repository of monitoring and response behaviors

Each of these three types of ontology has two independent ontologies, one for normal operations and one for crisis operations. Ontology merging is a core technical problem that has to be addressed.

Diagram 1:

The Adi structural ontology has three layers, as seen in Diagram 1. The first layer is a 8*4 framework developed through the study of natural language. The development of the Q framework is complex, as discussed in Part I. A more general Adi structural ontology exists for non-linguistic systems. In this case, the middle layer is called “systems or functional ontology”.

The second layer is derived from Q through a semi-formal method, as discussed in Part II. The semi-formal methods include the description and use of precedent rules, polarities, and control structures. In the linguistic system, the derivation of a functional or conceptual layer is also based on a study of co-occurrence patterns of elemental middle layer constructions called q-stems. These co-occurrence q-stem patterns are measured using the SLIP and Orb technologies (in the research environment created by Prueitt and Adi.) In the commercial software, Readware, the middle layer is a look up table consisting of 3618 functional elements (which can be called “concepts”).

The third layer re-applies a number of the semi-formal methods to the elements of the middle layer. In the linguistic system, the third layer is a cultural layer. This cultural layer reveals human interests and aspects of the cultural history as realized in legal theory and theory of social interactions. The discussion of the third layer is in Part III.

Diagram 2:

BCNGroup founders has suggested to industry three types of ontology for crisis management. This are listed above (Page 1). These ontology types each have two independent ontologies, one for non-crisis operation and one for crisis operation. Each of these ontologies must be developed so as to be completely interoperable in real time. See Ken Ewell’s note on RDF, topic maps and Readware ontology. It is also necessary that the entire ontology must be revealed inside of a communication system of the type proposed by Peter Stephenson at Eastern Michigan University. (see Architecture for Managing Incident Information )

The second type of ontology is an ontology derived from a real time conceptual mapping, sometimes called a conceptual role-up. The conceptual role-up can be done with at least three different commercial vendors, Applied Technical Systems Inc, Entrieva Inc and Readware Inc.

The third type of required ontology, for crisis management, is a workflow or process mapping ontology (see also “business rules”). These ontologies are sometimes developed using colored Petri nets, workflow software or a combination of the two. We will see that the Adi middle ontology has the possibility of providing a complementary modeling formalism for process mapping. The development of a technology bridge between work by Stephenson on modeling cyber attack trees using colored Petri nets will also be attempted this year.

Diagram 3:

In this diagram we suggest that physical infrastructure ontology can be encoded as an expression of the elementary constructions of the substructural ontology Q.

We also suggest that any workflow ontology can be encoded as a realization of the Adi cultural-environmental ontology.

The process by which the three types of ontology, required for crisis management, might be converted into a structured ontology requires the use of the notions from stratified theory and the Process Compartment Hypothesis. This activity is what the BCNGroup founding committee hopes will be funded over the next 12 months.

The core team has discussed the integration of the six recommended crisis management ontology into a structural ontology based on Orb encoding, and expressing each of the three crisis management ontology types in one very functional and optimal crisis management system. Some of the issues are outlined on the next page.

Outline of steps involving integration

1.0: The Physical infrastructure ontologies require a get deal of work, perhaps 30 – 40 people working three months. The details of what the infrastructures are likely already partially encoded into databases or perhaps even into a XML with a good organizational model. This may reduce the time to get sufficient data on an inventory of the items in the physical infrastructure. This physical infrastructure ontology can be encoded as RDF, with perhaps some OWL (or more specific some OIL) inference. The exact plan that we might have depends on what has already been done.

2.0: The development of the workflow and process model involves many of the cutting edge activities in business. Business rules, work flow and data mining are the types of activities that are occupying many teams working on different parts of the encoding of information about the critical infrastructure and how responses might be handled in a crisis.

3.0: There are not many commercial technologies that have targeted the development of a map of conceptual expression. This technology has made much advancement since the early 1990s when Oracle acquired Artificial Linguistics and over the next 8 years spent 60 M to integrate linguistic services into the Oracle process kernel. Then it was removed, not because it was not good, but because the salesmen thought that conceptual indexing was not easy to sell.

3.1: There is a community of perhaps 200 academics who used to receive funding from the TREC and MUC programs at NIST. Of these most have no real contribution to make other than to academic publications that are either not readable or do not address the real problems in “text understanding”. Their work is statistical and does not address the cognitive aspects. A few who were really good have turned their interests towards other things.

3.2: Applied Technical System, Entrieva and Readware are the exceptions to failures and avoidance of the text understanding problem. The activity that was funded in the 1980s and early 1990s has essentially dissolved.

3.2.1: I would hope that those who see this history differentially would communicate information to the BCNGroup founding committee.

3.2.2: The market has been understood within the business models of Applied Technical System and Entrieva in a way that has provided them income. But the capability to map the conceptual expression in real time is not in this business model. Of course not. The market has not supported this type of activity. Intelliseek Inc has almost completely forgotten what the issues are in understanding the social discourse.

3.2: One should ask why the linguist-based technologies have not been accepted in the marketplace. One answer is likely to be that the technology is complex. But this is not the most sever problem. There is a community that has solved many of the key issues, but this community has to compete with a larger funded community whose approach is poorly grounded in the natural sciences.

3.2.1: There is an unfair competition, between a funded and incorrect community and a non-funded and correct community,

3.2.2: There is the difficulty in explaining the nature of cognitive aspects to the public.

3.2.3: By linguistic we mean also approaches made by Ballard and others that go directly at the notion of information.