home *** CD-ROM | disk | FTP | other *** search
-
- ANTIC PRESENTS
-
- EXPERT OPINION
- (demo)
-
- Copyright 1986 Antic and Mind Soft
-
- EXPERT SYSTEMS -- A SMALL HISTORY AND EXPLANATION
-
- When we talk about an 'expert,' we mean an individual who is
- widely recognized as being able to solve a particular type of
- problem that most other people cannot solve nearly as
- efficiently.
-
- If you are sick, you will most probably go to see a doctor. After
- several questions and different tests, your doctor makes a
- diagnosis and presents a treatment. If your car doesn't start you
- will certainly call the mechanic to make it start.
-
- If you have to pay too much income tax, you will probably ask an
- accountant or a tax counselor for advice.
-
- All these experts you are consulting are considered persons who
- have a large amount of good domain-specific knowledge. As a
- result of their experience, they may permit us to efficiently
- solve our problems.
-
- Today, there exist computer programs called Expert Systems. These
- systems are able to behave like human experts in the domains
- developed for (medical, electronic, financial, etc...) Expert
- systems are among the first important results of Artificial
- Intelligence research.
-
- Is this the end of the experts?
-
- The goal of these programs is not to supplant human experts, but
- rather to supplement and to multiply them and to extend their
- capabilities by making their knowledge available for a larger
- number of persons.
-
- Expert systems are composed of a knowledge base that consists of
- facts and heuristics about a specific domain, of an inference
- engine that contains the inference procedures and the control
- strategies and of a base of facts that contains the
- specifications of a current problem and the results found by the
- expert system. Expert systems are totally based on knowledge and
- are characterized by their ease of use and by their possibility
- to explain how they find a particular result and why they ask a
- question at a specific moment but also by their capacity to give
- a well adapted solution on each case.
-
- The natural language interface makes expert systems friendly and
- powerful and allows to modify and to brush up easily the
- knowledge base. The order of which knowledge is entered into the
- system does not affect the results. In certain cases, the system
- may even find results without any precise description of the
- problem.
-
-
- HOW TO USE THE EXPERT OPINION DEMO
-
-
- MIND SOFT has created an expert system based on the philosophy
- that anyone should be able to become a "knowledge engineer."
-
- Our aim is to provide users with a tool that will solve problems
- based on choices, diagnosis, identification and, in general,
- expertise.
-
- Demonstration of the ST version.
-
- If you have a color monitor, please set it in medium resolution.
- You should have four files: EXPERT and BASE (main modules), and 2
- knowledge base examples (ANIMALS and TVSET).
-
- EXPERT
-
- This module contains the expert system. Expert Opinion provides 3
- modes of inference:
-
- - Deduction (forward chaining which goes from
- facts to conclusions)
-
- - Verify hypothesis (backward chaining which starts
- from a conclusion and goes to facts)
-
- - Expertise (forward and backward chaining at the
- same time)
-
- NOTE: This demo is only capable of displaying a 15-rule
- knowledge base. The real system is capable of a 500 rule base in
- a 512K system. To give you an idea of how large that is,
- Stanford University built an expert system called Mycin in the
- early '80s that diagnosed infectious diseases. That knowledge
- base had less than 300 rules in it, and it's diagnosis proved to
- be more accurate in most situations than a real doctors. Expert
- Opinion is potentially more powerful than Mycin, depending upon
- the knowledge put into the base.
-
- Double click on EXPERT.PRG to run the program. Then select, from
- the File menu, "Open knowledge base". Open TVSET to activate the
- knowledge base related to TV breakdowns. Now it is possible to
- use the three modes.
-
- 1) DEDUCTION
-
- This mode is to be used when the user is facing a problem for
- which he can describe what's going wrong. Basically, he should
- be able to formulate facts (or descriptions). From those
- formulated facts, Expert Opinion will infer deduction.
-
- Example: To formulate facts, select in the Inference Menu: "Add
- facts", and type 'NO IMAGE' (on the first line). Then click:
- "OK". Another way to do this is to add facts from a dictionary
- automatically generated by BASE (the knowledge base manager).
- This has two advantages: the user may refer to existing words and
- does not need to use the keyboard (mouse-controlled). For adding
- facts from dictionary, select, from Inference Menu: "Add facts by
- dictionary". If the fact to be input is 'THE SCREEN IS ON', you
- click on 'THE SCREEN' which will appear in the left bottom
- corner. For the next word, click 'IS' and last 'ON'. The fact
- 'THE SCREEN IS ON' is then composed. Click: "Store" and "OK".
- Now it is possible to make a deduction.
-
- Select "DEDUCE" in the Inference Menu. 1 new fact will appear. A
- deduction was thus made. It is now possible to return to the Base
- of facts (Edit Menu) and ask Expert Opinion how this new fact was
- deduced. After having selected Base of facts, the base will
- appear with 3 facts. You may then select the deduced fact by
- clicking on it (it will appear on last line) and then by clicking
- 'How ?' Expert Opinion will immediately explain to you how the
- fact was deduced by referring to a rule (contained in the
- knowledge base). It is then possible to check the mentioned rule
- by selecting: "Display a rule" in the Edit Menu.
-
- Important: before leaving the deduction mode, don't forget to
- erase the base of facts by clicking "Empty the base of facts" in
- Edit Menu.
-
- 2) VERIFY HYPOTHESIS
-
- To use this mode we will choose another knowledge base. First,
- close the TVSET base by selecting "Close" in File Menu. Then
- select: "Open knowledge base" in File Menu and select ANIMALS.
- Expert Opinion is now ready to answer questions about animals
- (although in this demo, there are only 15 rules instead of the
- maximum of 500). This mode is necessary when the user is able to
- formulate a hypothesis and wants to reach a specified goal.
- Select: "VERIFY a Hypothesis" in the Inference Menu.
-
- Example: type 'the animal is an elephant'. Then click OK.
- Questions will appear immediatly on the screen. The user may
- answer by Yes, No or No Idea if he is not able to make up his
- mind. At this stage it is possible to ask Expert Opinion to
- explain why he is asking such question. Try it by clicking on
- "Why?". The program will then justify its line of reasoning.
- Under the question a 'Why window' will appear with the
- explanation. 'Level 2' means that Expert Opinion is obliged to go
- through 1 sub-goal to verify the main goal : 'the animal is an
- elephant'. The first sub-goal is to prove that the animal is a
- mammal. If you want to discover the second sub-goal, click
- "Why?" until you reach level one. Click "OK" to return to the
- question. According to the user's answers, Expert Opinion will be
- able to verify (or not) the hypothesis. The system accepts up to
- 10 sub-goals.
-
- Important: a hypothesis must always correspond to the conclusion
- of a rule and not to one of the conditions composing the rule
- (see BASE). If you need help, click on Assistance when the system
- asks to give a hypothesis.
-
- 3) EXPERTISE
-
- This mode has been designed to allow a user who is not able to
- formulate facts or build a hypothesis, but wants a solution to
- his problem. In that case, this mode will do forward and
- backward chainings. That means that it will raise questions and
- will also generate new facts whenever possible. Expert Opinion
- will try to deduce facts based on the answers provided by the
- user. To operate this mode click Expertise in Inference Menu. A
- screen will appear with a maximum of 10 classes. In fact, when
- the expert has created his knowledge base he has been asked to
- split his knowledge in classes. For instance, a motor can have
- the following classes: brakes, clutch, electric system,
- carburator, ignition, etc. In this way, the user might be able
- to locate his problem in one class to enable Expert Opinion to
- ask questions related to that specific class of problem in order
- to avoid the user to have to go through the entire knowledge
- base. Expert Opinion will only ask questions concerning the
- rules contained in the class you request (unless you ask to be
- queried on the entire knowledge base).
-
- According to the users answers, Expert Opinion will develop a
- base of facts that should help the user to solve his problem.
-
- 4) BASE
-
- This module allows the user to write the rules which compose the
- base of knowledge. A rule is a proposition containing at least
- one condition and one conclusion, for instance: 'the animal is a
- mammal'. All the rules can be split into classes. Each rule can
- receive a certainty value, which is an indicative coefficient
- attributed to each rule in order to classify rules between
- themselves. This classification is interesting when using the
- inference engine because questions always begin with the highest
- certainties. The solution is a zone for special comments,
- warnings and solutions.
-
- The first thing to do when creating a knowledge base is to name
- your base (click "Create a new base" in File Menu). A mask will
- appear that has to be filled in by the expert. Each time the
- expert has finished writing one rule he must click Store or Quit.
- In the demo version of Expert Opinion, the BASE MANAGER is not
- capable of actually writing a working knowledge base to disk (we
- have to save something for the product, right?) Take a look
- through the commands of the Menu Bar and you will see the way the
- BASE MANAGER works. The Menu Facilities is very handly to
- provide the user with information concerning the meaning of each
- command.
-
- EXPERT OPINION WILL BE RELEASED MAY 1, 1986. WE INVITE YOU TO...
-
- JOIN THE KNOWLEDGE ENGINEERS!
-
- əəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəəə