home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Language/OS - Multiplatform Resource Library
/
LANGUAGE OS.iso
/
prolog
/
ai.prl
/
summers.lha
/
esdoc
< prev
next >
Wrap
Text File
|
1991-04-02
|
87KB
|
1,786 lines
ES Expert System
Copyright (C) 1989
Eric Summers
127 Hopeland Lane
Sterling, VA 22170
USER'S GUIDE
Version 1.0
October 1990
WARRANTY
The ES Expert System carries no warranty of any kind. In no event
shall Eric Summers be liable for any claims resulting from the use
or misuse of the ES Expert System. Any use of ES is a "use at your
own risk" endeavour.
REGISTRATION
To register as a user of ES expert system mail the following
information to me (Please Print):
Name ____________________________ Phone ______________________________
Address ______________________________________________________________
City ____________________________ State _____________ Zip ____________
Company _________________________ Phone ______________________________
Comments _____________________________________________________________
______________________________________________________________________
______________________________________________________________________
.pa
Chapter 1 - General
ES is a knowledge base development system designed to run on the
IBM/AT (or 100% compatible clones). ES was developed so IBM
microcomputer users can inexpensively experiment with using
expert system software.
1.1 SYSTEM REQUIREMENTS
IBM/AT or equivalent with 640KB of memory and a hard disk
(approximately 1 MB). A CGA monitor monitor or monochrome monitor
is required. It may be possible to use ES with floppy disks, but has
not been tested.
1.2 REPORTING BUGS
An attempt has been made to debug ES as completely as possible.
However, there are undoubtedly bugs remaining. I will attempt to fix
the bugs if you notify me. If you are the only person that finds a
given bug, I will put your problem on the bottom of my queue. If
you think you have a problem, write me and include a floppy disk
with the knowledge base, screen dumps, and any other evidence of the
problem. No specific dates can be given for fixing software bugs.
1.3 NEW FEATURES
If you can think of a new feature that ES should have, please write
me. If there is sufficient demand, I will consider developing a new
version of ES that includes your suggestion.
1.4 CONSULTATION
At this time, NO consultation is offered.
1.5 INSTALLATION
There are five files associated with the ES expert system, this file
ESDOC, the executable file ES.EXE, and three demonstration knowledge
bases: ES.KB, SPILL.KB and DISEASE.KB. You should make a directory,
call it EXPERT and copy all five files into this directory.
1.6 FEATURES
ES has the following features:
1. Backwards chaining, using global knowledge (non-frame based);
and Forward chaining. During forward chaining, backwards chaining
is automatically activated if any facts can be inferred.
2. Fact and rule knowledge editor. FIRST, create the facts then
create the RULES. The rule editor selects rule antecedents and
consequents from the list of facts.
3. Fuzzy set relations: fz=, fz<, fz> and fz<> are supported (see
Chapter 8). The "usual" expert system rule relations are also
supported: =, <>, <, <=, > and >=.
4. Certainty factors for both rule definition and user assertion.
CF's and fuzzy set inclusion factors are combined when a
consultation is done.
5. Semi-non-monotonic inference engine. Assertions for
conflicting inferences can be changed, depending upon fuzzy set or
rule CF's, ie, ES can change its mind regarding a truth that it
may have determined. However, ES is limited. No attempt is made
to analyze/reassert/retract any conclusions that may have
previously been determined from an inference who's truth CF is
subsequently reversed. If you need a comprehensive truth
maintenance capability, the KEE (tm) expert system shell by
Intellicorp does an excellent job of managing conflicts, using the
"worlds" concept (Patrick Winston discusses "worlds" in his book
"Artificial Intelligence"). ES also supports multiple lines of
reasoning.
6. ES rules are CONJUNCTIONS of ANTECEDENT SETs and multiple
consequents. Antecedent sets include:
* AND ANTECEDENT SETS - the set is true if ALL of the
set antecedents prove true. Resultant CF is the SMALLEST of
antecedent proved true in the set. An antecedent is the
smallest component of an antecedent set (eg color = black).
* OR ANTECEDENT SETS - the set is true if any one of
the antecedents proves true (ALL set antecedents are tested);
resultant CF is the LARGEST CF proved.
* XOR ANTECEDENT SETS - stop processing when the first
antecedent in the set proves true.
7. Knowledge explanation. ES explains the reasons for questions
as well as question and value explanations.
1.7 Limitations
The maximum size of a knowledge base (facts and rules) is
approximately 80K. If runtime errors occur, a knowledge base (KB)
that is too large may be the problem. For both the read-only and
read-write editors that ES provides, the maximum size that can be
edited is approximately 64K. When displaying inferences,
conclusion, and scenario, the maximum is also 64K.
1.8 Example Knowledge Bases
ES provides three simple knowledge bases:
1. ES.KB - this catchall KB demonstrates conflict resolution.
The fact "nice" can be proved via several rules, and demonstrates
the conflict resolution facility of ES. Several rules also make
use of fuzzy set relations.
2. SPILL.KB - this KB demonstrates identifying the chemical name
and contact for a chemical that may have been spilled. This is a
demonstration-only KB.
3. DISEASE.KB - this KB demonstrates identifying the disease name
from sickness attributes. This is a demo-only KB.
1.9 Menus and Cursors
Press the right and left arrow keys to move the cursor. To select a
pulldown option, press the Enter key (or type one of the highlighted
Capital letters). In most cases, ESC will return you to the next
level up in the menu hiearchy. When creating NEW knowledge (facts and
rules), ES will force you to complete the option you requested. If ES
cannot find the knowledge (fact, rules, etc.), the option will not
activate (eg, nothing will happen if you attempt to display fact
names, and there are no facts yet defined in the KB).
1.10 Editors
ES has two editors: one a read-write editor with control keys that
work much like WordStar (tm); the other editor is read-only, and the
only keys that work are the arrow and page keys. With either editor,
ESC or F10 is the exit key.
1.11 Expert System Theory
This user's guide contributes little towards the theory of expert
systems. However, an inexpensive book, "Understanding Expert
Systems", by Mike Van Horn, The Waite Group, Publ., is a good starting
point to learn about expert systems theory.
1.12 Suggestion for Getting Started
To get started using ES, I suggest that you experiment with the
SPILL.KB or DISEASE.KB knowledge bases. First, do backward and
forward chaining (Consult option) with both KB's to understand the
user aspect, then go onto the ES.KB knowledge base. Next, take the
SPILL.KB and expand it to recognize a new chemical that may have been
spilled. After this, you should be able to develop your own knowledge
bases.
1.13 Future?
If somehow able to acquire more advanced hardware than the IBM/PC, I
would like to be able to develop more capable versions of ES for other
platforms, including PS/2, Unix, and Apple Mac II. Some areas I would
like to investigate include:
* Developing a frame-based expert system with object
oriented capabilities including classes and subclasses,
instances, multiple inheritance, methods, daemons, and a
forward and backward chaining capability.
* ES curently has only a limited conflict resolution
capability. I would like to develop a more capable conflict
resolution capability that allows conflicts to be further
analysed, including removing inferences that become
subsequently invalid due to a conflict, and reassertion of
previously deleted conclusions due to a once-invalid
inference becoming valid again during the course of a
consultation.
* Developing a more capable user interface that makes use of
mouse, menus, windows, electronic forms, graphics and
hypertext/hypermedia.
* Developing the ability to imbed selected ES capabilities
in a user application. This will allow more intelligent and
easier to use applications to be developed.
* Developing the ability to define "hooks" for user-
developed programs, such as allowing a user to integrate
their own truth maintenance algorithm.
* Developing the ability to access information from databases
and other sources.
.pa
Chapter 2 - Main Pulldown Menu
This chapter explains how to use the Main Pulldown Menu. To run the
ES expert system type es and then press Enter. The following
pulldown menu will appear:
-----------------------------------------------------------------------------
| File fActs Rules Consult TruthMte fZsets |
-----------------------------------------------------------------------------
------------------------------------------------------------------
| Press arrows to move cursor. Select with CAP letter or Enter. |
------------------------------------------------------------------
Figure 2.1. Main Pulldown Menu
From the Main Pulldown Menu, type F to access the File menu which
appears below (Fig 2.2). Chapter 3 details use of the File menu.
-----------------------------------------------------------------------------
| File fActs Rules Consult TruthMte fZsets |
-----------------------------------------------------------------------------
|Load knowledge base |
|Save knowledge base |
|New knowledge base |
|Purpose for knowledge base|
|edit pUrpose for KB |
|eDit a disk file |
|temporary eXit to DOS |
|Quit |
----------------------------
Figure 2.2. File Options
From the Main Pulldown Menu, type A to access the Facts menu which
appears below (Fig 2.3). Chapter 4 details use of the Facts menu.
-----------------------------------------------------------------------------
| File fActs Rules Consult TruthMte fZsets |
-----------------------------------------------------------------------------
|Display fact names|
|Add new facts |
|Edit existing fact|
|deLete facts |
--------------------
Figure 2.3. Facts Options
From the Main Pulldown Menu, type R to access the Rules menu which
appears below (Fig 2.4). Chapter 5 details use of the Rules menu.
-----------------------------------------------------------------------------
| File fActs Rules Consult TruthMte fZsets |
-----------------------------------------------------------------------------
|Display rules |
|Add new rules |
|deLete existing rules|
|Edit existing rules |
|Rename rules |
-----------------------
Figure 2.4. Rules Options
From the Main Pulldown Menu, type C to access the Consult menu which
appears below (Fig 2.5). Chapter 6 details use of the Consult menu.
-----------------------------------------------------------------------------
| File fActs Rules Consult TruthMte fZsets |
-----------------------------------------------------------------------------
|Backward chain (determine goal) |
|Forward chain (reason data => conclusions)|
|display Conclusions |
|display Scenario |
--------------------------------------------
Figure 2.5. Consult Option
From the Main Pulldown Menu, type T to access the TruthMte menu which
appears below (Fig 2.6). Chapter 7 details use of the TruthMte menu.
-----------------------------------------------------------------------------
| File fActs Rules Consult TruthMte fZsets |
-----------------------------------------------------------------------------
|set Threshold CF |
|Display threshold CF |
|set for 'Running' TMS|
|set for 'Average' TMS|
-----------------------
Figure 2.6. Truth Maintenance Options
From the Main Pulldown Menu, type Z to access the fZsets menu which
appears below (Fig 2.7). Chapter 8 details use of the fZsets menu.
-----------------------------------------------------------------------------
| File fActs Rules Consult TruthMte fZsets |
-----------------------------------------------------------------------------
|New fuzzy sets |
|Delete fuzzy sets|
|Edit fuzzy sets |
-------------------
Figure 2.7. Fuzzy Set Options
.pa
Chapter 3 - File Menu
3.1 Loading a knowledge base
If you select L (Load Knowledge base) from the File menu, the
following menu will appear. Move the cursor to the knowledge base you
wish to load from disk, then press Enter.
-----------------------------------------------------------------------------
| File fActs Rules Consult TruthMte fZsets |
-----------------------------------------------------------------------------
---- Select knowledge base ----
| DISEASE.KB |
| ES.KB |
| SPILL.KB |
| ..\ |
| ANN\ |
| ANSWERS\ |
| BGI\ |
| EXAMPLES\ |
| FRAME\ |
| HYPTEXT\ |
| NUEXPERT\ |
| NUFRAME\ |
| OBJECT\ |
| PIE\ |
-------------------------------
Figure 3.2. Loading a Knowledge Base
3.2 Saving a knowledge base
If you select S (Save Knowledge base) from the File menu, the
following read prompt will appear. Type the name of the file and
press Enter. You must include a file type of KB. During KB creation
and editing, periodic saves are suggested.
---------------------------------------------------
| Type filename to store knowledge in: myfile.kb |
---------------------------------------------------
Figure 3.3. Saving a Knowledge base to disk
3.3 Making a New Knowledge Base
If you select N (New Knowledge base) from the File menu, the following
(yes/no) read prompt will appear. Type y if you wish to save the
current knowledge base; you will then be prompted for file name as in
section 3.2 above.
--------------------------------
| Save current knowledge (y/n) |
--------------------------------
Figure 3.4 New Knowledge Base
3.4 Purpose for Knowledge Base
If you select P (Purpose of KB) from the File menu, the following
read/only (R/O) editor screen will appear. Press ESC or F10 to exit
editor.
--------------------------------- Purpose -----------------------------------
|This knowledge base helps identify chemical spills. |
| |
| |
| |
| |
| |
-----------------------------------------------------------------------------
Figure 3.5. Purpose for Knowledge Base
3.5 Editing Purpose for Knowledge Base
If you select U (Edit purpose for KB) from the File menu, the
following read/write (R/W) editor screen will appear. Now, type the
text that explains the purpose for your knowledge base. Press ESC or
F10 to exit editor.
--------------------------------- Purpose -----------------------------------
| Line 1 Col 1 Indent Insert |
|This knowledge base helps identify chemical spills. |
| |
| |
| |
-----------------------------------------------------------------------------
Figure 3.6 Editing Purpose for Knowledge Base
3.6 Editing a Text File
If you select D (Edit text file) from the File menu, you will be
prompted (see Fig 3.7) for a file name and a read/write editor screen
will appear. Type the text you wish to save in a disk file. Existing
disk files may also be edited. Press ESC or F10 to exit editor. The
knowledge bases used by ES are all in text form, and you may edit the
structures. However, any modification may invalidate your knowledge
base, and it is suggested that you NOT edit any KB's.
--------------------------------------
| Type name of file to edit: myfile |
--------------------------------------
Figure 3.7 Editing a text file
3.7 Temporary exit to DOS
You may temporarily exit to DOS if you press D from the File menu. To
return to ES from DOS, type the DOS EXIT command.
3.8 Quit
When you are finished using ES, select Q from the File menu. The
following prompt will then appear. Type y if you wish to permanently
exit ES or n to continue using ES.
-------------------------
| Are you sure (y/n): y |
-------------------------
Figure 3.8 Confirmation to Quit Prompt
.pa
Chapter 4. Facts Menu
There are three types of facts:
* numeric facts - the value is always a number (integer or real);
A valid range may be specified for the fact.
* single-valued symbolic - the value is always a symbol.
* multiple-valued symbolic - the value may be more than one
symbol.
4.1 Displaying Fact Names
If you type D (Display fact names) from the Facts menu, the following
popup menu will appear. Press up/down arrows, PGUP, PGDN, HOME and
END keys to move through the menu. Press Enter to return to Facts
Menu. All other ES popup menus follow these key conventions.
----- Fact names ----
|Contact |
|chemical |
|chemical type |
|color |
|density |
|need to do cleanup |
|quantity |
|smell |
---------------------
Figure 4.1 Display fact names
4.2 Add a new fact to knowledge base
To add a new fact to KB, type A from the Facts menu, and the following
read prompt will appear. If the name you type already exists, you
will be notified.
---------------------------------
| Type name of new fact: color |
---------------------------------
Figure 4.2 Adding name of new fact
The following menu will then appear. Select the type that you wish
the fact to be.
---- Select type fact ----
|Symbolic single valued |
|symbolic Multiple valued|
|Numeric |
--------------------------
Figure 4.3 Selecting new fact Type
If you select N (Numeric), you will be prompted for the fact question
(see Fig. 4.9) and be returned to the Facts Menu. If you select S
(single) or M (multiple), you will be prompted for the possible values
for the fact shown in Fig 4.17. You will be able to add as many
values for the symbolic fact as you need (see Fig 4.18).
4.3 Editing Facts
If you press E from the Facts menu, the following menu will appear.
--- Select fact to edit -------
|Contact |
|chemical |
|chemical type |
|color |
|density |
|need to do cleanup |
|quantity |
|smell |
-------------------------------
Figure 4.4 Select Fact to Edit
Move the cursor to the fact you wish to edit and press Enter. If you
select a numeric fact (in this example "quantity"), the following menu
will appear:
---Select edit opt for: quantity---
|display Fact |
|Set Range |
|deleTe range |
|Question |
|reaSon |
|question eXplanation |
|display Conclusions |
|Add a conclusion |
|eDit a conclusion |
|deLete a conclusion |
|edit fallthrouGh text |
|delete fallthrougH text |
-----------------------------------
Figure 4.5 Numeric fact edit menu
If you select F (display Fact) from the edit opt menu, the following
will display. Press Enter to return to Edit fact menu.
--- Values for fact: quantity ---
|Type --> numeric |
|Range: 0 to 10000000 |
---------------------------------
Figure 4.6 Numeric Fact Information
If you select S (Set Range) from edit fact menu, ES will prompt you
for the lower and upper bound as in the following. In each case, type
the real number and press Enter. If you do not type a number, the
prompt will appear until you do.
-------------------
| Lower bound: 0 |
-------------------
Figure 4.7 Setting Lower Bound
--------------------------
| Upper bound: 55333000 |
--------------------------
Figure 4.8 Setting Upper bound
Press T if you need to delete the range. If no range is present in
the KB, during a forward or backward chaining, any value for the fact
will do.
Press Q if you wish to edit the fact question text, and the following
R/W editor window will appear.
-------------------------------- Question -----------------------------------
| Line 1 Col 1 Indent Insert |
|Quantity of chemical spilled? |
| |
| |
| |
| |
-----------------------------------------------------------------------------
Figure 4.9 Editing Fact question
Press R if you need to edit the reason for the fact question, and the
following R/W editor window will appear.
--------------------------------- Reason ------------------------------------
| Line 1 Col 1 Indent Insert |
|A large enough quantity is critical. |
|If sufficient, cleanup will be necessary. |
| |
| |
| |
| |
-----------------------------------------------------------------------------
Figure 4.10 Editing Reason for Question
Type S if you need to editing the explantion for the fact question,
and the following R/W editor window will appear.
---------------------------- Fact Explanation -------------------------------
| Line 1 Col 49 Indent Insert |
|Quantity of chemical spill, measured in gallons. |
| |
| |
| |
| |
-----------------------------------------------------------------------------
Figure 4.11 Explanation for Fact Question
From the Edit Fact menu, press C to display the text for a fact
conclusion, selecting specfic conclusion as in Fig 4.12.
------ Display which conclusion ---
|quantity = 0 |
|quantity = 99.99 |
-----------------------------------
Figure 4.12 Select Conclusion to display
From Fig 4.10 above, select the conclusion you wish to display, and
the following R/W editor window will appear (Fig 4.13).
-------------------------Conclusion=>quantity = 0----------------------------
|Strange, the quantity of chemical spilled was 0.0!!! |
| |
| |
| |
| |
| |
-----------------------------------------------------------------------------
Figure 4.13 Display specific conclusion text
If you need to edit a fact conclusion, select D from the edit optin
menu, and you can select the conclusion to edit from the menu in Fig
4.14. If you select a conclusion, a R/W edit window will appear and
you can editing the existing conclusion text.
------ Edit which conclusion ---
|quantity = 0 |
--------------------------------
Figure 4.14 Select Specific Conclusion to Edit
To delete a fact conclusion, select L from the edit option menu, and
the following menu will appear. Move the cursor to the fact you want
deleted and press Enter.
------ Delete which conclusion ---
|quantity = 0 |
----------------------------------
Figure 4.15 Delete specific conclusion text
Fall through - Fall through is a special condition test in ES.
It is impossible to have a rule for every condition in a KB, such
as color, density and smell of a chemical, which in the SPILL.KB
knowledge base, define what the chemical spill is. ES allows the
consultation (FC or BC) to stop if "fall through" is defined for
a given fact such as "chemical". You set this up by selecting G
(edit fallthrouGh text) from the symbolic or numeric fact edit
(Fig. 4.5 above and Fig. 4.16 below). A R/W window will then
appear and you can edit the text for the fall through condition.
This text is subsequently displayed (via selecting Conclusions
from Consult menu (Fig. 2.5 above) if during consultation, the
fact value cannot be inferred. Should you decide to remove fall
through for a fact, select H (delete fallthrougH text) from menu
in Fig. 4.5 above or Fig. 4.16 below.
If you select a symbolic fact to edit (in this example "color"), the
following menu will appear:
---Select edit opt for: color---
|Display values |
|Add a value |
|dElete a value |
|Question |
|reaSon |
|edit question eXplanation |
|edit Value explanations |
|display Conclusions |
|add a cOnclusion |
|ediT a conclusion |
|deLete a conclusion |
|edit fallthrouGh text |
|delete fallthrougH text |
--------------------------------
Figure 4.16 Editing Symbolic Fact
If you select A (Add a value), the following prompt will appear. Type
the new value for the fact.
---------------------------------------
| Type a value for fact color: orange |
---------------------------------------
Figure 4.17 New fact value Prompt
You will be allowed to add another value (Fig 4.18) as shown:
--- Add another value? ---
|Yes |
|No |
--------------------------
Figure 4.18 Add another new value Prompt
If you need to delete a fact value, select D from the Edit symbolic
fact menu (Fig. 4.16) and the following menu will appear. Move cursor
to the fact value you wish to delete and press Enter.
--- Delete value for: color---
|red |
|black |
|green |
|brown |
|yellow |
|clear |
|orange |
------------------------------
Figure 4.19 Delete Fact Value Menu
From Edit Fact menu (Fig 4.5 or 4.16) type Q if you desire to edit the
text for the fact question and the following R/W editor window will
appear. Correct the text and press F10 or ESC to exit.
-------------------------------- Question -----------------------------------
| Line 1 Col 1 Indent Insert |
|What color is the chemical? |
| |
| |
| |
| |
-----------------------------------------------------------------------------
Figure 4.20 Edit Fact Question Text
Type R (edit Reason) if you need to edit the reason for the fact
question (See Fig 4.21).
--------------------------------- Reason ------------------------------------
| Line 1 Col 1 Indent Insert |
|Color goes a long way towards identifying the chemical. |
| |
| |
| |
| |
-----------------------------------------------------------------------------
Figure 4.21 Edit Fact Reason Text
Type X if you need to edit the fact explanation (see Fig 4.22).
---------------------------- Fact Explanation -------------------------------
| Line 2 Col 40 Indent Insert |
|The major color of the chemical such as |
|red, green, blue, ... or a hue of such. |
| |
| |
-----------------------------------------------------------------------------
Figure 4.22 Fact Explanation Text
4.4 Delete Facts
If you need to delete a fact from the knowledge base, select L from
the Facts menu and the following menu (Fig 4.23) will appear. Select
the fact you need to delete and press Enter. A confirmation menu will
appear (Fig. 4.24).
--- Select fact to delete ---
|Contact |
|chemical |
|chemical type |
|color |
|density |
|need to do cleanup |
|quantity |
|smell |
-----------------------------
Figure 4.23 Delete specific Fact
-------------------------
| Are you sure (y/n): y |
-------------------------
Figure 4.24 Delete Fact Confirmation Menu
.pa
Chapter 5 - Rules Menu
After you have created some facts (see chapter 4), you can then create
and edit the rules, which make use of facts and fact values. This
chapter discusses the editing of rules.
5.1 Displaying Rules
From the Rules Menu, press D to display the current rules, and the
following screen will appear. Move the cursor until the name of the
rule you wish to display appears, then press Enter. The rule will
appear in the "Rule" window.
--- Select rule name to display --- ----------------- Rule -----------------
|Acid type rule | |Name: crude oil rule |
|Base type rule | |IF |
|Carroll rule | | AND SET: |
|Rule for gasoline | | density = high |
|Shirley rule | | smell = oily |
|Sodium hydroxide rule | | color = black |
|alcohol rule | | |
|benzene rule | |THEN |
|cleanup factor rule | | chemical = crude oil CF=0.8 |
|crude oil rule | | |
|petrol type rule | | |
|rule ammonia | | |
|rule hcl | | |
|rule plastic | | |
|sulfuric acid rule | | |
----------------------------------- ----------------------------------------
Figure 5.1 Display Rules
5.2 Deleting a rule
To delete an existing rule, select L from the Rules Menu, and the
following menu will appear (Fig 5.2). Move the cursor to the rule you
wish to delete and press Enter. A confirmation prompt will appear
(Fig 5.3).
--- Select rule name to delete ---- ----------------- Rule -----------------
|Acid type rule | |Name: crude oil rule |
|Base type rule | |IF |
|Carroll rule | | AND SET: |
|Rule for gasoline | | density = high |
|Shirley rule | | smell = oily |
|Sodium hydroxide rule | | color = black |
|alcohol rule | | |
|benzene rule | |THEN |
|cleanup factor rule | | chemical = crude oil CF=0.8 |
|crude oil rule | | |
|petrol type rule | | |
|rule ammonia | | |
|rule hcl | | |
----------------------------------- ----------------------------------------
Figure 5.2 Delete a rule
-------------------------
| Are you sure (y/n): y |
-------------------------
Figure 5.3 Delete Rule Confirmation
5.3 Adding a new Rule
To add a new rule, select A from the Rules menu and the following
prompt will appear. Type the name of the new rule. You will be
notified if the name already exists.
-------------------------------------------
| Type name for new rule: crude oil rule |
-------------------------------------------
Figure 5.4 Add new rule name
Next you will be prompted (Fig 5.5) for the antecedent set type.
--- Select type antecedent set ---
|And |
|Or |
|Xor |
----------------------------------
Figure 5.5 Select Set Type
ES will then build the individual antecedents for the antecedent set
(See Fig. 5.6).
--- Select antecedent name: ---
|Contact |
|chemical |
|chemical type |
|color |
|density |
|need to do cleanup |
|quantity |
|smell |
-------------------------------
Figure 5.6 Select antecedent name
If you select a fact that is symbolic, the following menu will appear.
Select the appropriate antecedent relation (Fig 5.7). If the fact
selected is numeric, the menu will have addtional relations (<, <=, >,
>=).
--- Select symbolic operator: ---
|= |
|<> |
---------------------------------
Figure 5.7 Select operator
Next, select a value for the relation, shown in Fig. 5.8.
--- Select value for color---
|white |
|red |
|black |
|green |
|brown |
|yellow |
|clear |
-----------------------------
Figure 5.8 Select antecedent Value
If you select a numeric fact (Fig 5.6) the following type menu will
appear (Fig 5.9).
--- Select numeric type: ---
|Real number |
|fuzzy Set |
----------------------------
Figure 5.9 Select numeric antecedent value type
If you select Real number from Fig. 5.9, the following numeric
relations will appear (Fig 5.10)
--- Select numeric operator: ---
|= |
|<> |
|< |
|<= |
|> |
|>= |
--------------------------------
Figure 5.10 Real number relations
You will then be prompted for the number value for the relation as in
Fig. 5.11.
---------------------------------------------
| Type a numeric value for: quantity 543210 |
---------------------------------------------
Figure 5.11 Real number value for relation
If you select fuzzy Set (see Fig 5.9), the following menu will appear,
select the fuzzy set relation.
--- Select fuzzy set relation ---
|fz= |
|fz<> |
|fz> |
|fz< |
---------------------------------
Figure 5.12 Select fuzzy set relation type
Next, select the fuzzy set symbol as in Fig 5.13.
--- Select fuzzy set symbol: ---
|large spill |
--------------------------------
Figure 5.13 Select fuzzy set symbol for antecedent
If more antecedents are needed for the antecedent set, select Y from
the following menu (Fig 5.14).
--- Another antecedent? ---
|Yes |
|No |
---------------------------
Figure 5.14 Add another antecedent
If you select N from the menu in Fig 5.14, the following menu will
appear, allowing you to create another antecedent set. If you select
Y, you will be prompted for another set and antecedent values.
--- Another antecedent set ---
|Yes |
|No |
------------------------------
Figure 5.15 Add another Ante Set
If you select N to the menu in Fig 5.15, you will be prompted for rule
consequent values, as in Fig. 5.16.
--- Select consequent name: ---
|Contact |
|chemical |
|chemical type |
|color |
|density |
|need to do cleanup |
|quantity |
|smell |
-------------------------------
Figure 5.16 Specific consequent name
When you select a consequent, from above, you will need to set the
value for the consequent, shown in Fig. 5.17.
--- Select value for chemical---
|crude oil |
|HCl |
|NaOH |
|H2SO4 |
|Plastic |
|Benzene |
|gasoline |
|alcohol |
|ammonia |
--------------------------------
Figure 5.17 Select consequent value
Next, set the certainty factor for the rule consequent, shown in Fig
5.18. Type a value between -1.0 and 1.0.
-----------------
| Type CF: 1.0 |
-----------------
Figure 5.18 Specify consequent CF
If your rule needs to prove more than one consequent, select Y from
the following menu (Fig 5.19).
--- Another consequent ---
|Yes |
|No |
--------------------------
Figure 5.19 Add another consequent
If you select N from the above menu, you will be given the opportunity
to create another new rule (Fig 5.20).
--- Add another rule ---
|Yes |
|No |
------------------------
Figure 5.20 Add another Rule
WHen finished creating rules (You select N to menu in Fig. 5.20), you
are returned to the Rule Menu.
5.4 Renaming rules
To rename a rule, select R from the Rules Menu and the following menu
will appear (5.21). Move cursor to the rule you need to rename and
press Enter.
--- Select rule name to rename ---
|Acid type rule |
|Base type rule |
|Carroll rule |
|Rule for gasoline |
|Shirley rule |
|Sodium hydroxide rule |
|alcohol rule |
|benzene rule |
|cleanup factor rule |
|crude oil rule |
|petrol type rule |
|rule ammonia |
|rule hcl |
|rule plastic |
|sulfuric acid rule |
----------------------------------
Figure 5.21 Select rule to Rename
The following prompt will next appear. Type the new name of the rule
and press Enter. You will be notified if rulename already exists.
----------------------------------------------------------------
| Type name to replace crude oil rule with: my crude oil rule |
----------------------------------------------------------------
Figure 5.22 Supply rule name
5.6 Cloning a new rule
To clone a rule, select C from the Rules Menu, and the following menu
will appear. Select the current rule you need to clone and press
Enter.
--- Select rule name to clone ---
|Acid type rule |
|Base type rule |
|Carroll rule |
|Rule for gasoline |
|Shirley rule |
|Sodium hydroxide rule |
|alcohol rule |
|benzene rule |
|cleanup factor rule |
|crude oil rule |
|petrol type rule |
|rule ammonia |
|rule hcl |
|rule plastic |
|sulfuric acid rule |
----------------------------------
Figure 5.23 Select rule to Clone
You will need to give the cloned rule a new name, as in Fig. 5.24.
---------------------------------------------------
| Name for rule you just cloned: my ammonia rule |
---------------------------------------------------
Figure 5.24 Supply cloned rule name
5.7 Editing existing rules
If you need to edit an existing rule, select E from the Rules Menu,
and the following menu will appear (Fig 5.25).
.pa
--- Select rule name to edit ---
|Acid type rule |
|Base type rule |
|Carroll rule |
|Rule for gasoline |
|Shirley rule |
|Sodium hydroxide rule |
|alcohol rule |
|benzene rule |
|cleanup factor rule |
|my ammonia rule |
|my crude oil rule |
|petrol type rule |
|rule ammonia |
|rule hcl |
|rule plastic |
|sulfuric acid rule |
--------------------------------
Figure 5.25 Select rule to edit
After the desired rule is selected, the following screen will appear
(Fig 5.26).
--- Edit rule component: ----------------------------- Rule -----------------
|Edit antecedent set | |Name: my crude oil rule |
|add new Antecedent set | |IF |
|add new Consequent | | AND SET: |
|Delete consequent | | color = black |
---------------------------- | |
|THEN |
------------ Antecedent Set: -----------| chemical = crude oil CF=1 |
| || |
| || |
| || |
| || |
| || |
| || |
| || |
| || |
| || |
| || |
| || |
| || |
| || |
-----------------------------------------------------------------------------
Figure 5.26 Edit Rule Components Menu
If you desire to edit the antecedent sets, select E from the above
menu (Fig 5.26). and the following menu will appear (Fig 5.27).
.pa
--- Select antecedent option: ------------------------ Rule -----------------
|Add antecedent to antecedent set | |Name: my crude oil rule |
|Delete antecedent from antecedent set| |IF |
|dElete complete antecedent set | | AND SET: |
|Next antecedent set | | color = black |
--------------------------------------- | |
|THEN |
------------ Antecedent Set: -----------| chemical = crude oil CF=1 |
| AND SET: || |
| color = black || |
| || |
| || |
| || |
| || |
| || |
| || |
| || |
| || |
| || |
| || |
| || |
-----------------------------------------------------------------------------
Figure 5.27 Edit Antecedent Set
Select A to add an antecdent to the antecedent set shown in the lower
left window. You will be prompted for antecedent values as in Figures
5.6, 5.7, 5.8, 5.9, 5.10, 5.11, 5.12 and 5.13. If you wish to delete
an antecedent from the antecedent set shown above (Fig 5.27), press D.
You can select the specific antecedents to delete (See Fig 5.28).
--- Delete antecedent --------------------------------- Rule ----------------
|density = high | |Name: crude oil rule |
|smell = oily | |IF |
|color = black | | AND SET: |
|quantity fz= large | | density = high |
------------------------- | smell = oily |
| color = black |
------------ Antecedent Set: -----------| quantity fz= large |
| AND SET: || |
| density = high ||THEN |
| smell = oily || chemical = crude oil CF=0.8 |
| color = black || |
| quantity fz= large || |
| || |
| || |
| || |
| || |
| || |
| || |
| || |
| || |
| || |
| || |
-----------------------------------------------------------------------------
Figure 5.28 Delete specified Antecedent Set antecedents
If you need to delete the complete antecedent set shown, press E from
the menu as in Fig. 5.27. To go to the next antecedent set for the
rule that is being edited, press N (Note that the ES rule editor
"cycles" (N option) through the antecedent sets from first to last).
After the last antecedent set appears, you will be returned to the
Edit Rule Components screen as in Fig. 5.26.
If you need to Add a new consequent to the rule, press C from the Edit
Rule Components screen (Fig. 5.26) and the series of menus (Fig. 5.16,
5.17, 5.18 and 5.19) will prompt you for the new rule consequents.
If you need to delete a consequent from the rule, press D from the
Edit Rule Components screen (Fig. 5.26) and the following menu will
appear. Move cursor to the rule consequent you wish to delete and
press Enter.
--- Delete which consequent? ---
|chemical = crude oil 0.8 |
|need to do cleanup = yes 0.345|
--------------------------------
Figure 5.29 Delete rule consequent
.pa
Chapter 6. Consult Menu
6.1 Backward Chaining
If you select B from the Consult menu, the following menu will appear.
Move the cursor to the goal you wish to determine, and press Enter. ES
will attempt to determine the name of the goal, and any conclusions as
a result of proving the goal. This is done by asking you questions.
---- Determine which goal ----
|Contact |
|chemical |
|chemical type |
|need to do cleanup |
------------------------------
Figure 6.1 - Select Goal for Backward chaining
ES will ask you questions to determine the requested goal. You will
see the following type of screen will appear.
----------------------------- Question ------------------------------------
|What color is the chemical? |
| |
| |
| |
| |
| |
---------------------------------------------------------------------------
--- Select option --- ----------------- Rule -----------------
|Assert a value | |Name: sulfuric acid rule |
|Unknown | |IF |
|Conclusions | | AND SET: |
|Explanations | | color = yellow |
|sTop consultation | | density = medium |
|eXit expert system | | smell = sour |
--------------------| |THEN |
| chemical = H2SO4 CF=1 |
| |
| |
| |
| |
----------------------------------------
Figure 6.2 - Question menu
If you do not know what the fact value is, select Unknown (U). You
may also display any Conclusions (C) already determined, Explanations
(E) for the Question or any parameters, stop (T) the consultation or
exit (X) to DOS.
Now, select A (Assert a value) and the following will appear:
----------------------------- Question ------------------------------------
|What color is the chemical? |
| |
| |
| |
| |
| |
---------------------------------------------------------------------------
--- Select value for fact --- ----------------- Rule -----------------
|white | |Name: sulfuric acid rule |
|red | |IF |
|black | | AND SET: |
|green | | color = yellow |
|brown | | density = medium |
|yellow | | smell = sour |
|clear | | |
----------------------------- |THEN |
| chemical = H2SO4 CF=1 |
| |
| |
| |
| |
----------------------------------------
Figure 6.3 Assert a fact value menu
Now select the appropriate value for the fact. You will next be
prompted for the CF for the fact value, shown as:
---------------------------------------------------------
| Type CF (-1.0 <= CF <= 1.0) for [color = black]: 0.75 |
---------------------------------------------------------
Figure 6.4 Set CF for a fact value
Depending on how the rules are networked, you may be prompted for the
values for several facts.
ES also allows you to assert numeric and multi-valued facts. Menus
and explantions are similar. When asserting a numeric fact, you
simply type the number that you know the fact is. When asserting a
multiple-valued fact, you select as many values as you know are
applicable to the fact. You will be prompted for the CF for each fact
value asserted.
6.2 Forward Chaining
Forward chaining is triggered by selecting F from the Consult menu.
If F is selected, the following menu will appear, allowing you to
assert the value(s) for selected facts.
--- Select fact to assert ---
| Contact |
| color |
| density |
| need to do cleanup |
| quantity |
| smell |
-----------------------------
Figure 6.5 Assert fact value for Forward Chaining
ES will fire all rules to determine any conclusions. ES will
automatically backward chain if it is possible to infer a fact value.
6.3 Conclusions
After you are finished with the Forward or Backward chaining, select C
from the Consult menu, and the conclusions determined will be
displayed, one at a time as shown below. Note that if a specified
fact value cannot be inferred, and "fall through" is defined for that
fact, the fall through text will be displayed.
------------------------------ Conclusion -------------------------------
|chemical = crude oil CF=0.8 |
| |
|The chemical spilled is crude oil. It is very |
|dangerous to the environment. It's destruction |
|can last for years. |
| |
| |
-------------------------------------------------------------------------
Figure 6.6. Conclusion with Text
Keep pressing ESC or F10 to display all conclusions. If no text is
defined for a conclusion, the following conclusion window will appear:
--------------- No conclusion text defined for inferences: ------------------
|color = black CF=1 |
|density = high CF=1 |
|smell = oily CF=1 |
|chemical type = petroleum CF=0.8 |
| |
| |
-----------------------------------------------------------------------------
Figure 6.7 Conclusions that do not have Text
6.3 Scenario
To display in the R/O editor any rules that were fired, select S from
the Consult menu, and the following will appear; Press ESC or F10 to
exit the editor.
------------------------------ Scenario: --------------------------------
|Firing rule ==> sulfuric acid rule ... |
| Proved: color = black CF= 1 |
| Failed: color = yellow |
| Rule: sulfuric acid rule failed..... |
|Firing rule ==> Shirley rule ... |
|Firing rule ==> Base type rule ... |
|Firing rule ==> crude oil rule ... |
| Proved: density = high CF= 1 |
| Proved: smell = oily CF= 1 |
| Proved: chemical = crude oil CF= 0.8 |
| Rule: crude oil rule proved true..... |
|Firing rule ==> rule hcl ... |
| Failed: color = green |
| Rule: rule hcl failed..... |
|Firing rule ==> benzene rule ... |
| Failed: color = brown |
| Rule: benzene rule failed..... |
|Firing rule ==> Rule for gasoline ... |
| Failed: color = red |
| Rule: Rule for gasoline failed..... |
|Firing rule ==> Sodium hydroxide rule ... |
| Failed: color = white |
-------------------------------------------------------------------------
Figure 6.8 Rule Firing Scenario
.pa
Chapter 7. Truth Maintenance Menu
ES supports two simple truth maintenance paradigms:
* Running - the resultant CF is determined by the following
formula whenever the certainty of an inference differs in the rule
base: CF = CF1 + CF2*(1 - CF1), eg, if color = red with rule #1
determined CF = 0.8 and color = red with rule #2 CF = 0.2, the
resultant CF = 0.8 + 0.2*(1 - 0.8) = 0.84.
* Average - the resultant CF is the average of the CF's
determined either by inference or by assertion, eg, if color = red
with CF=0.8 is determined by rule #1 and color = red with CF=0.4
is determined by rule #2, the resultant CF = (0.8 + 0.4)/2 =0.6.
7.1 Setting a threshold CF
For backwards or forward chaining, the default threshold certainty
factor is 0.0, which says that a fact can be determined as "true" if
its CF is at or above the threshold CF. However, you may wish to make
the threshold higher, say 0.1, so that a fact is proved true only if
its CF is above "unknown" (CF =0.0). To reset the threshold CF,
select T from the TruthMTE menu, and the following will appear.
--------------------------------------------------
| Type new threshold CF (-1.0 <= CF <= 1.0): 0.1 |
--------------------------------------------------
Figure 7.1 Setting threshold CF
7.2 Displaying Current Threshold CF
Press D from the TruthMte menu to display the current Threshold CF
value. The following prompt will appear. Press Enter to continue.
-----------------------------------------------
| Threshold CF= 0.1, press Enter to continue: |
-----------------------------------------------
Figure 7.2 Current threshold value
7.3 Display current TMS Type
Select either R (Running) or A (Average) to set the current TMS.
Consult (Backward or Forward chaining) will make use of the current
TMS when doing conflict resolution.
7.4 Display current TMS
From the TruthMte menu, press C to display the current TMS. The
following will appear:
------------------------------------------
| Current TMS= 'Average'. Press Enter: |
------------------------------------------
Figure 7.3 - Current TMS
Chapter 8. Fuzzy Sets Menu
This chapter discusses fuzzy sets and fuzzy relations and the ability
for ES to create and edit fuzzy sets.
8.1 Overview of Fuzzy Sets
A brief overview of fuzzy set relations follow. An exhaustive study
of fuzzy sets is given in "Fuzzy Mathematical Techniques with
Applications" by Abraham Kandel, Addison Wesley, Publ. See chapter 1
which defines fuzzy sets.
8.2 Example
An example is age => middle-aged. This is certainly subjective, but a
person could be considered middle aged between 40 and 60, but some
might consider 38 or 62 as middle-aged, but with less certainty.
Fuzzy set age fz= middle_aged in the ES.KB knowledge base shows an
example, described as:
30 -1.0 (not member fuzzy set)
38 0.2 (slightly member fuzzy set)
40 1.0 (full member fuzzy set)
60 1.0 (ditto)
70 -1.0 (definitely not member fuzzy set)
NOTE: Values of fuzzy set inclusion factors that are not
explicitly given are extrapolated by ES to determine the CF. Note,
that we are also synonimously using CF and "inclusion factor".
You might want to make use of fuzzy set relations in your knowledge
base. Consider developing three fuzzy sets (assuming sex = male):
1. weight fz= heavy - 150 (-1.0) is not heavy, 200 is
somewhat heavy (maybe 0.2), 300 is certainly heavy (1.0)
2. height fz= tall - 7 feet (1.0) is certainly tall, 6 feet
(0.2) is somewhat tall, and 5 feet (-1.0) is certainly not
tall.
3. girth fz= sizeable - a person with a 30 inch waist is not
sizeable (-1.0), 35 (0.1?) and 50 (1.0).
A rule that determines if a person is a large person could look like:
RULE LARGE PERSON
IF
OR SET:
weight fz= heavy
height fz= tall
girth fz= sizeable
THEN
person = large person CF= 1.0
If you are backward chaining (Consult/Backward Chain), and the goal is
person, ES will prompt you for the weight, height and width and make
use of the largest inclusion (CF) factor and thus determine if the
person under consideration actually is a large person. If you are the
owner of a Large Men's Store, you would be very interested in the
above rule.
8.3 Fuzzy Relations
The following fuzzy set rule relations are supported:
* fz= -- if the relation is a member of the fuzzy set (CF > 0.0),
the rule antecedent proves true
eg: age fz= middle_age.
If age = 39, CF = 0.9
FZ=: CF (X) must be > 0.0 AND within envelope
**********************************************************
* CF ^ *
* | . . *
* | . . *
* | . . .. *
* | . . . . *
* | . X . . X . *
* 0.0 ---------------------------------------> VALUE *
* . . . *
* . . . *
**********************************************************
* fz< -- if the relation proves less than the smallest inclusion
factor of a fuzzy set.
eg: age fz< middle_age
If age = 32, CF = -1.0
FZ<: CF (X) must be less than any number contained in envelope
**********************************************************
* CF ^ *
* | . . *
* | . . *
* | . . .. *
* | X . . . . *
* | . . . . *
* 0.0 ---------------------------------------> VALUE *
* . . . *
* . . . *
**********************************************************
* fz> -- if the relation proves greater than the largest factor
in a fuzzy set.
eg: age fz> middle_age
If age = 70, CF = -1.0
FZ>: X (CF) must be greater than any number contained in envelope
**********************************************************
* CF ^ *
* | . . *
* | . . *
* | . . .. *
* | . . . . X *
* | . . . . *
* 0.0 ---------------------------------------> VALUE *
* . . . *
* . . . *
**********************************************************
* fz<> -- proves trues if the number is NOT a member of a fuzzy
set.
eg: age fz<> middle_age.
If age = 30, CF = 1.0
FZ<>: X (CF) must NOT be contained within envelope
**********************************************************
* CF ^ *
* | . . *
* | . . *
* | . . .. *
* | X . . X . . X *
* | . . . . *
* 0.0 ---------------------------------------> VALUE *
* . . . *
* . . . *
**********************************************************
8.4 Adding a New Fuzzy Set
If you press N from the fZsets menu, the following will appear:
--- Select numeric fact ---
|age |
|height |
|salary |
|weight |
---------------------------
Figure 8.1 - Adding New Fuzzy Set
You should type the name of the fuzzy set as follows:
-----------------------------------------------------------
| Type fuzzy set relation[name] for fact age: very short |
-----------------------------------------------------------
Figure 8.2 - New fuzzy set name
Next, the following will appear. Type a numeric value for the fuzzy
set:
---------------------------------------
| Type value for age = very short: 4 |
---------------------------------------
Figure 8.3 - Type value for fuzzy set
Next, type the inclusion factor for the value:
--------------------------------------
| Type CF (-1.0 <= CF <= +1.0) : 1.0 |
--------------------------------------
Figure 8.4 - Type inclusion factor
ES allows as many fuzzy values (and associated inclusion factors) as
necessary. After a value/inclusion factor is input, the following
menu will appear. Select Y if more value/inclusion factors need to be
defined.
--- Add another fz set component: ---
|Yes |
|No |
-------------------------------------
Figure 8.5 - Add another fz set value/inclusion factor
8.5 Delete fuzzy sets
To delete a fuzzy set, select D from the fZsets menu, and the
following menu will appear. Move cursor to fuzzy set you wish to
delete and press Enter.
--- Select numeric fact ---
|age |
|height |
|salary |
|weight |
---------------------------
Figure 8.6 - Select numeric fact
If you select "* next fuzzy set" from the menu (Fig 8.7), ES scrolls
through the available fuzzy sets of the fact you selected. ESC at
this point returns you to the fZsets pulldown (Fig 2.60).
-----Fact age: very short----
|* delete complete fuzzy set|
|* next fuzzy set |
|4 1 |
-----------------------------
Figure 8.7 - Select fuzzy set
If you select "* delete complete fuzzy set", the following
confirmation prompt appears. Type y to delete fuzzy set.
-------------------------
| Are you sure (y/n): y |
-------------------------
Figure 8.8 - Confirmation delete fuzzy set prompt
8.6 Edit Fuzzy Sets
Press E from the fZsets menu, and the following menu will appear.
Move cursor to fact who's fuzzy sets you wish to edit and press Enter.
---Edit which fz set for: age---
|13 |
|16 |
|40 |
|59
|age extremes |
|middle_aged |
|old |
|very_old |
|very_young |
|young |
--------------------------------
Figure 8.9 Select fuzzy set to edit
Move the cursor to the fuzzy set you wish to edit and press Enter.
The following menu will then appear:
-------age : middle_aged-------
|* add fuzzy set components |
|* delete fuzzy set components|
|* plot fuzzy set |
|0 -1 |
|20 -0.95 |
|32 -0.2 |
|40 1 |
|59 1 |
|62 -0.1 |
-------------------------------
Figure 8.10 - Edit fuzzy set options
If you select "* add fuzzy set components" option, you will be
prompted for the fuzzy set component and inclusion factor (see Section
8.4 above, with figures 8.2, 8.3, 8.4, and 8.5).
If you select "* delete fuzzy set components" option from fig. 8.10,
the following menu will appear. Move cursor to component to delete
and press Enter. Press ESC key to stop deleting components.
---Delete fz set component for: age middle_aged---
|0 -1 |
|20 -0.95 |
|35 -0.2 |
|38 0.1 |
|40 1 |
|59 1 |
|60 0.5 |
|65 -1 |
--------------------------------------------------
Fig. 8.11 - delete specific fuzzy set components
To plot the specific fuzzy set, select "* Plot fuzzy set". A full
screen graphic will appear, similar to plots described in 8.3 above.
When finished viewing the graphic, press Enter.
