Manual for Clips is available at Clips homepage

An intelligent computer program that uses knowledge and inference procedures to solve problems that are difficult enough to require significant human expertise for their solutions.

–Prof. Edward Feigenbauma, Stanford University

In particular, expers systems can be compared to a program that simulates work of an human expert in decision making tasks. An Expert is a person who has very specialist knowledge, unavailable for majority of people. An Expert can solve a problem that most of people would not be able to solve, or would not do it effectively enough.

Experts systems using Clips can be created in two ways:

• in console (extremally inconvinient)

• in a file, which is later loaded into the Clips interpreter with load command.

An exerise is aimed at creating simple calculator for creditworthiness. Rules presented below are fictional

Creditworthiness will be defined as an ability of a borrower to pay current debt in a timely manner.

• CLIPS inherited partially syntax from LISP, hence all commands have to be ebraced wtih parentheses.

• To start Clips console, type in the Unix console

  clips

• To exit Clips type

  (exit)

• To load a file with rules/facts type

  (load filename)
  
  (reset)

  Command (reset) loads facts defined with deffacts (See sectopnFacts).

• To clear knowledge base type

  (clear)

• To run inference process type

  (run)

• To launch help, type

  (help)

• Comments in Clips are denoted with ; (semicolon)

• It is common that Clips files have .clp extension.

1. Facts are defined with deffacts command. For example to add several facts that would describe person name, last name, age, etc. following code should be placed in a file or typed in the Clips console:

   (deffacts my_facts "comment"
   
       (name jan)
   
       (surname kowalski)
   
       (age 24))
   

   To add defined facts the (reset)) comman has to be invoked. Attention Facts within deffacts are not connected with each other in any way. In an example from above to a fact base three independent facts would be added: (name jan), (surname kowalski) and (age 24).

1. Facts can be added to a knowledge base with assert command:

   (assert (age 23))
   

   The assert command would add a fact immediatelly, howewer after invoking (reset) command, only facts defined with deffact would remain and those added with assert will be lost.

1. Facts can consist of one or more parts:

   (deffacts myFacts
   
     (year 2010)
   
     (person jan kowalski 45)
   
     (informatyka stosowana))
   

1. Removal and modification of facts is available with retract and modify commands

1. To list all facts from knowledge base, following command should be used:

   (facts)

Using deffacts, define facts that would represent persoal data such as name, surname, age, salary.

Load created file into the Clips, and list all files.

1. To define rules defrule keyword is used. To add arule which will print Hello world message, if there is name jan fact in fact base, following code should be used:

   (defrule rule-1 "Comment"
   
     (name jan)
   
   =>
   
     (printout t "Hello World!" crlf))

1. If there is more than one fact in a knowledge base that starts with name we may want to print a name of the person to whom the salutation is addressed. In this case, the rule form example above should look like that:

   (defrule rule-1 "Comment"
   
     (name ?value)
   
   =>
   
     (printout t "Hello World!" ?value crlf))

1. To check if a value is equal to, less than, or greater than the other value a test functino is used. For exampel to check if the age of a person is less than 18 following rule should be defined:

   (defrule rule-1 "Comment"
   
     (name ?imie)
   
     (age ?age)
   
     (test (< ?age 18))
   
   =>
   
     (printout t "Hi " ?imie ". Your not supposed to be here!" crlf))

Define two rules that would describe weather a person can be a potential debtor. If an age of a peerson is less than 18 it can not be, if it is geater and the person's salary exceedes 1500$ - a credit can be granted.

After the inference process is run, „Credit granted” or „Credit not granted” message should be printed in the screen.

A fat can contain any number of elements. Both following facts are valid:

(assert (name szymon))

(assert (person with a name szymon))

Aobove notation is not convinient for facts that sould represent more complex data.

for more complex facts, deftemplate lub defclass constructs are used:

(deftemplate person

  (slot name)

  (slot surname)

  (slot age))

It is possible to acces some of the slots ommiting other:

(assert (person (name jan) (surname kowalski) (age 50))



(defrule rule ""

  (person (age ?value))

  (test (>= 18 ?value))

 =>

  (printout t "Hello!" crlf))

Define a temaplate person which would store data defined in previous exercises. Modify defined rules to work with template data.

Modules in Clips are used fro several reasons

1. They help to manage large knowledge bases by groupping rules into smaller sets

1. They improve efficiency (for each module, a separate RETE network is created)

1. They allow for simple inference control (focus command)

A module is defined with a defmodule construct:

(defmodule myModule)

To define that some rule belongs to some module a:: operator is used (it is simiellar to known from C++ scope resolution operator):

(defrule myModule::rule ""

   =>

   (printout t "---- Rule in myModule ----" crlf)

While defining modules it can be defined if facts added by rules/templates/classes that belongs to the module should be visible outside its scope:

(defmodule myModule (export ?ALL))

For exported data to be visible inside other module, it has to be specified which data and from which module should be imported::

(defmodule myModule (export ?ALL)

(defmodule myOtherModule (import myModule ?ALL))

All data not explicetely assign to any module by defauld belongs to a MAIN module.

It is required sometimes to redefine MAIN module to export all data, so it can be imported in other modules:

(defmodule MAIN (export ?ALL))

it is possible to define order in which defined modules will be processed by an inference engine.

1. An arbitrary order can be defined:

   (defrule MAIN::startrule ""
   
      =>
   
      (printout t "---- STARTING ----" crlf)
   
      (focus salutation greeting userMessage)	; inference control
   
   )

1. Modules can be switched in Right Hand Side (RHS) of rule:

   (defrule MAIN::reugla ""
   
     (osoba (wiek ?value))
   
     (test (>= 18 ?value))
   
    =>
   
     (printout t "I'm switching from MAIN to adoult" crlf)
   
     (focus adoult))

Define three modules: PrivateCapability FinancialCapability and FinalCapability. Add facts privCap, finCap, endCap that would be set in aforementioned modules.

privCap should be calculated based on table below:

finCap should be calculated based on table below:

endCap should be calculated according to the following equation:

(zdpriv*2 + zdfin*5)*1000.

Modules should be fired in following order (use focus):

1. PrivateCapability

1. FinancialCapability

1. FinalCapability

To read data from the jeyboard a (read) command is used:

  defrule regula ""

=>

  (printout t "Podaj swoj wiek: ")

  (bind ?wiek (read))

  (printout t "Podaj swoje imie: ")

  (bind ?imie (read))

  (printout t "Podaj swoje nazwisko: ")

  (bind ?nazwisko (read))

  (assert(osoba(imie ?imie)(nazwisko ?nazwisko)(wiek ?wiek))))

Reading data from the user can be done only in RHS of the rule.

Modify program from previous exercise by providing user interface that would allow user to type his famili status, age, salary.

After that an inference process should be fired and information about the amount of credid allowd for the user printed.

• kilka słów wprowadzenia teoretycznego

• więcej przykładów „za ršczkę”

• w materiałach wykłšd w pdf do pobrania

Z braku lepszego miejsca tutaj studenci wpisujš komentarze natury ogólnej do tego lab.