ExperTalk

Features

In Expertalk you get:

• A rasoning engine based on Facts and Rules, see examples
• There are 2 rasoning engines
  • Backward chaining (a demonstrator)
  • Forward chaining
• You can trace the rasoning so that you can see how ExperTalk operated to get the result
• You can have at any time an explanation of how any fact in the factsbase has been produced, at the whole deep level or at the level.
• In the rules you can specify Universal quantifiers, see examples
• Any Smalltalk object can be included in Facts
• With the DBF access you can load the factsbase directly from .DBF file data creating any facts you need from fields in the DBF file, that are really common in PC applications (optional)
• There is a Interactive mode that allow expertalk to be used with incomplete factsbase, and that will ask the user to confirm required facts it needs to demonstrate a fact, this will prompt the user to answear the truth of the fact ExperTalk need, here the user can answear Yes, No, a sure factor, or inspect the facts and rulebase, asking for explanations, list of facts, list of rules, why expertalk is asking the fact, and abort the rasoning process.
• Every fact have an associated Sure factor that allow you to deal with uncertanty, this means that not only true and false facts can be used but every fact have its own sure factor that will be propagated in the rasoning process. Sure factors are respresented by Fuzzy logic value
• ExperTalk is loop proof, it recognize when a demonstration cannot be produced because it will generate an infinite loop and will cut that branch of the rasoning automatically, so data engineers need not to worry about the consistency of all the rules.
• ExperTalk can access Smalltalk, executing blocks or compiling at runtime sources of blocks populated with metavariables, and executing them. Glossary
• ExperTalk can be easily customized and integrated with an existing application.
• A User interface is supplied to use it standalone or to inspect factsbase, you can use more than one factsbase at a time, The user interface allow you to simply create facts, load and save facts and/or rules, load data from a DBF, and to run the rasoning engines to demonstrate facts.
• Create facts is really easy and the end user can easily create them if you are not using it with special objects.
• There is no functor to be used, but every element of the fact can be used as a functor or as an attribute. The unification mechanism will find the matching facts on the base of what is supplied.
• Simple facts and rules can be esily read from strings.

Examples

Facts

Rules

Backward chaining

Forward chaining

Explanation

Trace

Glossary

• Fact: is the base of the rasoning process, can be thought, and is, like a sentence, i.e.: "John is brother of Susan", any Smalltalk object can be an element of a fact, and facts can be element of facts, creating a nested structure, each fact has an sure factor associated that is a fuzzy logic value
• Rule: is like: "if A & B then C & D", it is composed of antecedents the "if" part and consequents the "then" (more than one), they will be used recursively or iteratively by the rasoning engines to drive the rasoning process and produce new facts. i.e.: "if John is brother of Susan & Megan is doughter of Susan then John is uncle of Megan". In rules you can also use metavariables (universal quantifiers) that allow you to sepcify more generic rules that will apply on the factsbase i.e.: "if ?A is brother of ?B & ?C is son of ?B then ?A is uncle of ?C"
• Fuzzy logic value: is a value in the fuzzy logic arithmetic, it is like a probabilty factor, but operation does not behave like probability that cannot be used on incomplete databases, a fuzzy logic value is in the range 0-1 and logical operation like: and, or, not, ... are computed differently than in the probability.
• Universal quantifier, Metavariable: is a variable, that can assume values in the rasoning process, normally values of the variable will be loaded from the factsbase and populated in the rule, in ExperTalk a metavariable is identified beacuse it is a string that start with a question mark, i.e.: ?PERSON ?A ?B are metavariables or universal quantifiers, you can discover anything eated by John demonstarting a fact like: "John eats ?WHAT", in the rules metavariables define general rules, a metavariable is identified by the name and its scope is only in the rule (or in the fact to demonstrate) same variable name in different rules, are considere different
• Backward chaining: is the rasoning process that goes backward to see if an asked fact is true (producable) by the expert (set of facts and rules) so it start from the given fact and see if it is true, if not look for all the rules that can produce the fact and try to verifythe rules
• Forward chaining: is the process of producing new facts appliing the rules, every rule will be evaluated and if all the antecedents (the if part) are verified (also instantiating metavariables) all the consequents (the then part) will be added to the factsbase
• Factsbase: the set of fact that the expert knows
• Rulesbase: the set of rules that the expert knows

Add a User

Download
• Manual
• ExperTalk You will need a user and password to download and extract the files

  See also: Nik Smalltalk Objects for Download

  Nicola Zordan