Search the Internet to find a knowledge-acquisition development of Web-based expert systems. How

2. Search the Internet to find a knowledge-acquisition development of Web-based expert systems. How

tool that was not introduced in this chapter and sum-

are Web -based systems different from traditional marize its major functional features.

systems?

• G R O U P PROJECT Development a Small Rule-Based Expert System

pie. Many qualitative factors influence the choice. for Wine Selection

Delegating the decision to a waiter can be risky in some

Building a small knowledge base with an expert system

restaurants and even riskier at home if you are trying to

shell is fairly easy. Here is an example. Selecting an

impress someone. So it may be useful for a restaurant to

P A R T IV INTELLIGENT DECISION SUPPORT SYSTEMS

the C O R V I D expert system shell for this student project, ple, it is known that the wine depends on the menu selec- follow the steps specified below to develop your ES for

tion: You might have meat, fish, poultry, or pasta, and you wine selection:

may have an appetizer as well. Each fact-gathering state- ment is called a qualifier. For example, qualifier 1 is the

1. Specify the problem (pairing wine and food at a meat menu selection. Literally its text is the phrase "The restaurant).

meat menu selection is." Qualifier text generally ends in a 2. N a m e the system (Sommelier).

verb. The qualifier can assume the following values: (1) 3. Write the starting text ("Sommelier will help you

prime ribs, (2) grilled steak, and (3) filet mignon. These select a wine ...").

are called the values of the qualifier. 4. Decide on an appropriate coding for an uncertainty

N o w the first rule might be constructed as situation (e.g., 0/1,0 to 10, or 2100 to 100: we will use

OtolO). IF the meat menu selection is prime ribs 5. Decide on any other inferencing parameters as

T H E N pinot noir, confidence 9/10 required by the shell (e.g., use all the rules or only

A N D merlot, confidence 8/10 some of the rules; set the threshold levels).

A N D aged cabernet sauvignon, confidence 6/10. 6. Prepare any concluding note that you want the user

to see at the end of the consultation ("Thank you for This means that each of the three choices is appro- priate, but pinot noir is the best choice (9), followed by

using Sommelier. B o n appetit!"). merlot (8), and then by aged cabernet sauvignon (6). 7. Work with an expert to list the potential choices. To build this rule, create a new qualifier and then There are 12 possible wines, including aged cabernet

sauvignon blanc, crisp chardonnay, gerwiirztraminer, select the number of the qualifier (1) and the appropriate value (1 for prime ribs). This forms the IF part of the rule.

merlot, pinot noir, zinfandel, and so on. Click to the T H E N part of the rule and call up the 8. D e v e l o p a working knowledge of food and wine (this

choices (step 7) and select the appropriate values for the is the best part for the knowledge engineers who

T H E N part with the appropriate confidence level (don't must gain some familiarity with the problem

forget the confidence level or the system will recommend domain).

no wine, leading to a Tather dour evening). In this case 9. D e v e l o p a knowledge map with an expert (OK,

the three wines are appropriate, but pinot noir is the best maybe we did some testing on our own) in tabular

recommendation.

form. This helps us focus on the knowledge. When all the rules are applied to the particular meal, 10. Build the initial set of rules, the knowledge base, in

the system might not recommend pinot noir overall. As the shell, and debug them. In Exsys, this involves other rules are consulted, the certainty in the choice may developing qualifiers, which are the questions to be

decrease. For example, the appetizer and the cost of the posed to derive facts, and the answer set of possible wine may influence the recommendation. Another con- fact values for each one. Qualifiers are explained

below. sideration is the universe of the system's knowledge. If your menu selection is not on the considered list, the sys-

11. Demonstrate the system to the expert, refine the tem will not know what to recommend because the main rules, and collect additional cases and rules. Repeat

course is not part of its knowledge universe. steps 9-11 until the recommendation given by the Construct the knowledge base and implement the expert(s) and the system are the same. prototype to show some sample sessions. 12. Deploy the system ( B o n appetit!).

Source: Lisa Sandoval, a graduate student at California Write the rules in the standard format required by the

State University, Long Beach, developed this ES student shell. C O R V I D uses the concept of qualifiers. For exam-

project: "Your Sommelier."