A.J. Rhem & Associates

RHEM Knowledge Acquisition Framework

Concept

Task

Problem Solving Method

Result

Determine Domain Area

Talk to the domain expert and find out what problems or aspects of the problem solving process may be amenable to automation. Perform evaluation analysis to determine if the domain area is likely to use a Knowledge Base System.

Perform evaluation analysis using the Checklist Approach.

A qualified domain area(s) for the construction of a Knowledge Base System.

Decompose the Knowledge

Identify the type of knowledge being gathered (i.e. tacit, explicit, declarative, procedural). Break the domain into smaller subtasks. If the knowledge is contained in a database apply Data Mining techniques to extract the knowledge.

Several methods can be used here. To gather tacit knowledge the use of Decision Trees is recommended. To obtain knowledge from databases, the use of data mining techniques are highly desirable. Each method will facilitate the knowledge being broken into smaller more manageable subtasks.

The knowledge of the domain area is broken into smaller subtasks.

Determine Interdependencies

Analyze how individual components of knowledge are related and integrated when they are used to solve a task. Point out missing pieces of knowledge. Determine what pieces of knowledge are related and how. Detect inconsistencies among the various aspects of the knowledge.

Incorporate the use of Interdependency Models (IM). Tools such as EMeD (Expert Method Developer) allows users to specify problem solving knowledge and identifies the interdependencies between the various aspects of the knowledge.

Missing pieces of data are filled in, inconsistencies are corrected and a more complete view on how the knowledge components are related is established.

Recognize Knowledge Patterns

Identify patterns within the various types of knowledge gathered. Apply the knowledge gathered to similar situations that the domain expert may have encountered or to similar cases (Case Based Reasoning) that may have been solved previously.

Using Case Based Reasoning, developing decision trees and/or data mining techniques to discover patterns is the recommended approach.

The knowledge becomes more consistent and the knowledge base is able to be constructed more efficiently.

Determine Judgments in Knowledge

Determine if the knowledge being captured is judgmental (uncertain or "fuzzy") in nature. A determination must be made if there are conflicts between rules.

If we have conflicts between rules, we should ask the expert for an explicit, relative judgment between the active rules in that context.

Conflicting rules are identified and/or conflicting knowledge about an area of a domain is identified.

Perform Conflict Resolution

If the knowledge being captured has uncertainty or is "fuzzy" in nature the first step in resolving this uncertainty is to specify preconditions in the contexts of one or more of the conflicting rules to prevent them from being considered.

The effect of this is to add expertise, restricting the number of situations in which the rules are applicable.

Conflicts between rules and/or conflicts between knowledge are resolved.

Construct Knowledge Base

Perform knowledge modeling of the domain utilizing one or more of the techniques mentioned in the previous concepts.

Construct Knowledge Models utilizing the Unified Modeling Language (UML)

Knowledge Based System/Knowledge Management System is constructed for the domain area(s) under consideration.

Concept	Task	Problem Solving Method	Result
Determine Domain Area	Talk to the domain expert and find out what problems or aspects of the problem solving process may be amenable to automation. Perform evaluation analysis to determine if the domain area is likely to use a Knowledge Base System.	Perform evaluation analysis using the Checklist Approach.	A qualified domain area(s) for the construction of a Knowledge Base System.
Decompose the Knowledge	Identify the type of knowledge being gathered (i.e. tacit, explicit, declarative, procedural). Break the domain into smaller subtasks. If the knowledge is contained in a database apply Data Mining techniques to extract the knowledge.	Several methods can be used here. To gather tacit knowledge the use of Decision Trees is recommended. To obtain knowledge from databases, the use of data mining techniques are highly desirable. Each method will facilitate the knowledge being broken into smaller more manageable subtasks.	The knowledge of the domain area is broken into smaller subtasks.
Determine Interdependencies	Analyze how individual components of knowledge are related and integrated when they are used to solve a task. Point out missing pieces of knowledge. Determine what pieces of knowledge are related and how. Detect inconsistencies among the various aspects of the knowledge.	Incorporate the use of Interdependency Models (IM). Tools such as EMeD (Expert Method Developer) allows users to specify problem solving knowledge and identifies the interdependencies between the various aspects of the knowledge.	Missing pieces of data are filled in, inconsistencies are corrected and a more complete view on how the knowledge components are related is established.
Recognize Knowledge Patterns	Identify patterns within the various types of knowledge gathered. Apply the knowledge gathered to similar situations that the domain expert may have encountered or to similar cases (Case Based Reasoning) that may have been solved previously.	Using Case Based Reasoning, developing decision trees and/or data mining techniques to discover patterns is the recommended approach.	The knowledge becomes more consistent and the knowledge base is able to be constructed more efficiently.
Determine Judgments in Knowledge	Determine if the knowledge being captured is judgmental (uncertain or "fuzzy") in nature. A determination must be made if there are conflicts between rules.	If we have conflicts between rules, we should ask the expert for an explicit, relative judgment between the active rules in that context.	Conflicting rules are identified and/or conflicting knowledge about an area of a domain is identified.
Perform Conflict Resolution	If the knowledge being captured has uncertainty or is "fuzzy" in nature the first step in resolving this uncertainty is to specify preconditions in the contexts of one or more of the conflicting rules to prevent them from being considered.	The effect of this is to add expertise, restricting the number of situations in which the rules are applicable.	Conflicts between rules and/or conflicts between knowledge are resolved.
Construct Knowledge Base	Perform knowledge modeling of the domain utilizing one or more of the techniques mentioned in the previous concepts.	Construct Knowledge Models utilizing the Unified Modeling Language (UML)	Knowledge Based System/Knowledge Management System is constructed for the domain area(s) under consideration.