What should be an expert system?

Question

Programming

There was an idea for "training the brain", I want to write an expert system using the accumulated knowledge and experience. The question is not that very relevant and necessary, but from the fact that there is not a very large choice, either paid applications or free ones from the 80s of the last century and, to put it mildly, with not the best implementation. But I would not want to create it for a "horse in a vacuum", I need advice, for example:

1. how do ES see themselves in the understanding of an ordinary user, not just a developer?
   
2. What needs to be implemented in it?
   
3. What areas of application would it be better to adapt it to?
   

Answer 1

Apparently your competence is not in the field of ES, but in the field of programming. In that case I will try to prompt on ES.
Formally speaking (there is in every book on ES in the first chapter) an expert system is a ready-made system for which knowledge has been collected, their engineering and implementation, etc., according to the process. And for development, they use Expert System Shell - an ES shell. Hereinafter, I mean the shell everywhere.
Step one is to find out what exists. for this I advise you to study Jess in the book jess in action. jess is a much better implementation of CLIPS, plus using java adds more battery.
Then look at JBoss DRools, it has more functions, in particular backward chaining (this is also in Prolog, but there is no forward chaining there)
Previously, the Rete algorithm was not implemented everywhere, but now it seems to be in all normal ECOs.
Step two is to see what can be added. There is a book by Peter Jackson, Introduction to Expert Systems, which describes many functions from experimental or scientific expert systems, and rarely more than two are implemented in one ES shell. They can be added.
There are enough of them, and if not, you can ask.
Now the answers, in reverse order:
3. You can adapt to everything for which there is expert knowledge (well, or "expert" - that's not the point. You can start and get used to it - in any area. The main desire and energy)
But, for example, how they began to use ES now .
They began to be used in bureaucracies to replace bosses and other checkers from the bureaucracy with a system with rules tuned to check pieces of paper. more precisely, the questionnaires are entered electronically by the operator girl, and the ES checks them. up to checking whether it is possible to give a loan, etc.
In general, effective MBA managers have taken the microscope of computer scientists from the 80s and used it for their own purposes, hammering in a couple of copper nails, in the form of saving on bureaucracy.
2 "What needs to be implemented in it?" - something that is missing in the selected shell of the ES. maybe already there, looking at the task. for automation and replacement of bureaucrats with a mechanical bureaucratic machine, everything is already there, it seems.
Or maybe something is missing for the task. or the algorithms are slow. If it's a good challenge, maybe it will. The inventor of the Rete algorithm, for example, made new versions paid and protected by patents.
1. "how do ES see themselves in the understanding of an ordinary user, not just a developer?"
For an example above.
If visually, then the operator sees the fields for entering the fields of the questionnaires or the windows for a specific operation and at the end - ding - a message like "surname not entered" or conditionally, maximum, "do not issue a loan" or let's say some other machinery that requires intervention in reality, such as request a signature, or a client's passport.
And the one who writes the rules sees either text fields as a programmer, or a "more visual" type Query builder-a in MS Access, for the same rules.

Answer 2

Expert systems of the 80s - digitized sets of "if-then" rules.
This concept did not justify - I believe, because of the large number of "intermediate" entities that (and their relations) needed to be formalized, but it did not work out well.
That is, Watson is largely from there, but in order to get the point out of it, it took hiilliard man-years.
1. ordinary users are not in the subject
2. the language of the description of the rules
3. the conquest of the universe

Answer 3

Try to implement ES for diagnosing something, for example, diagnosing diseases or breaking any equipment. You can't get away from "If-then" - after all, you need to somehow analyze the object under study. You also need to create a knowledge base in which certain answers to certain conditions are already set. You can feel the logic on the service - Jin, who guesses any person and very successfully.