Brought to you by EarthWeb
IT Library Logo

Click Here!
Click Here!

Search the site:
 
EXPERT SEARCH -----
Programming Languages
Databases
Security
Web Services
Network Services
Middleware
Components
Operating Systems
User Interfaces
Groupware & Collaboration
Content Management
Productivity Applications
Hardware
Fun & Games

EarthWeb Direct EarthWeb Direct Fatbrain Auctions Support Source Answers

EarthWeb sites
Crossnodes
Datamation
Developer.com
DICE
EarthWeb.com
EarthWeb Direct
ERP Hub
Gamelan
GoCertify.com
HTMLGoodies
Intranet Journal
IT Knowledge
IT Library
JavaGoodies
JARS
JavaScripts.com
open source IT
RoadCoders
Y2K Info

Previous Table of Contents Next


3.5. CREATION OF THE DYNAMIC SUBMODEL

An analysis of the tasks obtained in the analysis stage makes it possible to outline the expert's dynamic or process model and check that there are not too many errors or omissions. Here, it is important to recall each task studied during the strategic knowledge identification stage. Then, we have to define a hierarchy between the tasks. For example, if task 1 has to fire task 2 to get information, then task 1 must be a parent of task 2, as shown in Figure 6.


FIGURE 6 Task hierarchy.

Then, the formalisms needed to execute the task have to be established. For this purpose, it is important to note down the possible goal values M = m1, etc. These should normally have been ascertained in the first stage of tactical knowledge identification, but they should be discussed with the expert for confirmation. Then, they need to be represented, as shown in Figure 7, where M represents the goal defined in TASK 1 of Figure 6, and m1, m2, and m3 are alternative subgoals characterizing the composition of goal M. Finally, A1 and A2 are the attributes concerning goal m1, and are elemental either because they are easily obtained from an inference or because they are explicit. That is, these attributes and/or their values are obtained either from an external source, such as a text or database, or are requested from the expert and/or users who enter a value or use a default value.

Then, meta-knowledge has to be considered. The best means of doing this is to (1) add the information to be shown during the inference to Figure 7 (for example, if A1 = a1.val1, the system must show the user a text on laws) and (b) implement priorities. If an attribute has a particular value, the system must immediately switch the process to a given attribute or ignore another one (for example, when A1 = a1.val2, check whether M = m1 is true; if not, check whether M = m2 is true, as shown in Figure 8).


FIGURE 7 Assignation of values to task goals.


FIGURE 8 Priority implementation.

Having identified the subdecisions (i.e., the individual stages leading up to the decision), it is important to get the expert to confirm that they are correct. At the same time, the expert is asked to state which concepts and attributes are used in each stage. This knowledge has to be recorded by (1) adding new metarules; (2) applying the subdecisions in the model and cross-checking that the model contains all the concepts required to carry out each stage; and (3) listing the order of the subdecisions in the process document.

Finally, the whole process should be completely and consistently established using a standard representation as shown in Table 5.

3.6. KNOWLEDGE MAP GENERATION

Whereas the dynamic model is composed of strategies, priorities, etc., making up the process submodel, and the static submodel contains the concept dictionary and its associated Concept/Attribute/Value table, the knowledge map is the synthesis of these two submodels. Although it basically represents the static part of expert knowledge, it should really also show its dynamic part.

A knowledge map, hereafter referred to as KM, is similar to a mental map. It is a method of representing the connections made by the brain when it understands facts about something in 2-D on paper. This map is easily recalled by the designer when revised some time later. Therefore, a KM represents the process of inferring attribute values. Relations between attributes and inferred values constitute an important part of knowledge. Liebowitz (1989) points out that KMs have several notable features. First, when fully developed, they are considered a powerful means of showing how much and what type of expert knowledge has been gathered. Second, the KM's visual format helps the KE in preparing later knowledge acquisition sessions, as it guides KE planning and behavior in these sessions. Without this positive and efficient expert-driven control mechanism, knowledge acquisition sessions would become random circumlocutions around subjects, that is, the knowledge acquisition sessions would deteriorate, becoming costly and excessively time consuming. Third, subjects can be embedded in the KM, reflecting the layers of knowledge gathered from the expert. This makes it possible to specify (using suitable signs, follow-on topics, following keywords) how to hierarchically order subjects to carry out a critical control function. Additionally, other references are permitted, for example, numbering knowledge acquisition sessions to allow direct backtracking of all the previous analytical steps until one can establish exactly where the knowledge was obtained and under what circumstances. Backtracking is required to keep adulteration of expert knowledge to a minimum. Finally, this notation can be transformed almost directly into a computer implementation. So, this approach allows complex elicited expert knowledge to be applied and implemented in a verifiable, documentable, and auditable manner.


TABLE 5
Standard Description of Procedures
 
KNOWLEDGE BASE NAME:
PURPOSES:
INFORMATION REQUIRED:
ACTIONS:
BASIC REASON:
INTERVIEW NOTES:
STATUS:

Additionally, it is important to note that a KM is not a semantic net. So, each attribute appearing in the left-hand side of a rule, for example, must appear in the KM connected to an attribute inferred in the right-hand side of the rule. But each connected attribute does not necessarily appear in each rule used to infer the value of the attribute in the right-hand side of the rule. The rules containing heuristic knowledge or meta-knowledge often only contain one or two attributes in the left-hand side of the rule.

The signs found when the expert and the KE jointly review the KM have two purposes. First, when the sign is followed by a keyword, the expert is asked to explain it. If the response is "No further knowledge and/or data or news are required," this is stated in the KM, no further inquiries are made, and the issue is considered closed. If, on the other hand, the response is affirmative, the expert has to be asked punctually and rapidly for some keys on this issue in later knowledge acquisition sessions. Similar questions are raised in the lower layers of the KM. In this manner, it is the expert who has final control over the depth to be achieved in a given subject, rather than the KE deciding which subjects to pursue and which not to pursue. So, the expert always controls the relative importance of each subject and how much attention is paid to it.


Previous Table of Contents Next

footer nav
Use of this site is subject certain Terms & Conditions.
Copyright (c) 1996-1999 EarthWeb, Inc.. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of EarthWeb is prohibited. Please read our privacy policy for details.