CommonKADS Expertise 1

CommonKADS Expertise

Presented by: Aly Wael Aboul Nour

Supervised by: Dr. A. Rafea

Introduction:

Expert Systems solves problems. Using a coherent framework for description of problem solving allow us to compare different expert systems.

CommonKADS that presented is a modeling framework that covers all aspects of knowledge-based application.

CommonKADS models

Organization model is the tool to analyze an organization, task model captures the global task within the organization and deal with certain functions including assignment of tasks to agents. The agent models describe agent capabilities, the expertise model agent competence involved in realizing the overall task. The communication model describe the communication behavior among agents.

We are concerned with the expertise model because it is the most one that involves Knowledge base systems

Knowledge Level:

Newell mentioned that there exists a level lying immediately above symbol level which is characterized by knowledge as the medium and the principle of rationality as law of behavior.

The concept is taking in CommonKADS as it

is the appropriate level for modeling the competence of knowledge-based systems. It calls for the description of problem solving behavior at a conceptual level that is independent from representation and implementation decision.

The knowledge level hypothesis does not give much guidance regarding how an intelligent agent can be described in knowledge level terms. An extension is needed , an agent has to impose a structure on his knowledge ascribing a particular role to different components. The agent will act rationally within this structure leads to a class of behavior deals with pragmatic and epistermogical problems.

e.g. MYCIN uses heuristic classification and knowledge is rational appropriate but achieve goals of the system.

Principle of differentiated rationally:

There is a question of why knowledge is structured in a certain way to comply with predetermined pragmatic and epistemic of task environments.

The differentiation known knowledge configuration and knowledge application (two step rationality).

The first part of differential rationality is concerned with modeling principles (rationalize model of expertise), the second part rationalize control on knowledge application.

Expertise Modeling Frame work:

In Common KADS theory we categorize knowledge level as:

  • Application knowledge
  • Problem Solving knowledge

Expertise Model Framework

Problem Solving knowledge is knowledge on problem solving methods in general and on strategic knowledge (when to use what type of solving methods)

Application knowledge contains facts about domain and control knowledge describing how the application task is achieved

  • Task knowledge captures goals of agents and the activities of achieving the goals
  • Inference knowledge describes the usage of domain knowledge in performing tasks via small reasoning steps
  • Domain knowledge gives vocabulary of the application domain.

Domain knowledge

Domain knowledge is a selection of all statement about the domain that together present a coherent view of the domain

Inference Knowledge

In

ference knowledge Are functional components defining basic reasoning steps operating on restricted parts of knowledge

Domain knowledge play roles in reasoning in:

  1. Static roles : points to domain knowledge elements that are used but not affected
  2. Dynamic roles: point to domain knowledge that is manipulated

Task Knowledge

Task definition describes goal of a task, its input and output roles, and their relation

Task body describes how the goal can be achieved by giving sub-goals assumptions and a task expression describing how the task goal can be achieved

Task decomposition tree illustrates in a graphical way which subtasks are to perform for performing the main tasks. Leaves are inferences

Complaintcausal-modelstatehas-manifestation

Generic Tasks:

Generic tasks of Chandrasekaran can used actually to instantiated problem solving methods applied to generic task definition at various level of grain size.

e.g. task specification can modeled as generic task (task definition and task body)

Conclusion:

A careful study of various knowledge modeling approaches reveals that different approaches are not incompatible, even though their terminology is different.

CommonKADS expertise is a solution framework to the expert interaction between several levels of knowledge thus we can introduce a dynamic methodology of providing a solution rather than first generation expert systems that interact with only frames, slots without maintaining a structure to solution providing.

Refereneces:

-Bob Wielinga, Walter Van de Velde, Guus Schreiber and Hans Akkermans, Towards a Unification of Knowledge Modeling Approaches.

-Christian Lockenholf, Siemens AG, CommonKADS, “Principles of Knowledge Modeling” ,

-Christian Lockenholf, Siemens AG, 1993, CommonKADS, “Structure of expertise Model”.