Expert System Shell (Java)

A simple and user-friendly expert system shell implemented in Java. The rule engine also support rule files written in Javascript.

Note that this expert system shell do not require external dependencies for its logics

Features

• Forward Rule Chaining
• Backward Rule Chaining
• Backward Rule Chaining with Prompt
• Support rules file written in Javascript

Install

Add the following dependency into your POM file:

<dependency>
  <groupId>com.github.cschen1205</groupId>
  <artifactId>java-expert-system-shell</artifactId>
  <version>1.0.1</version>
</dependency>

Usage

Add rules and initialize the rule engine

Below is an example to create a rule engine from scratch with a set of rules in java

private RuleInferenceEngine getInferenceEngine()
{
    RuleInferenceEngine rie=new KieRuleInferenceEngine();

    Rule rule=new Rule("Bicycle");
    rule.addAntecedent(new EqualsClause("vehicleType", "cycle"));
    rule.addAntecedent(new EqualsClause("num_wheels", "2"));
    rule.addAntecedent(new EqualsClause("motor", "no"));
    rule.setConsequent(new EqualsClause("vehicle", "Bicycle"));
    rie.addRule(rule);

    rule=new Rule("Tricycle");
    rule.addAntecedent(new EqualsClause("vehicleType", "cycle"));
    rule.addAntecedent(new EqualsClause("num_wheels", "3"));
    rule.addAntecedent(new EqualsClause("motor", "no"));
    rule.setConsequent(new EqualsClause("vehicle", "Tricycle"));
    rie.addRule(rule);

    rule=new Rule("Motorcycle");
    rule.addAntecedent(new EqualsClause("vehicleType", "cycle"));
    rule.addAntecedent(new EqualsClause("num_wheels", "2"));
    rule.addAntecedent(new EqualsClause("motor", "yes"));
    rule.setConsequent(new EqualsClause("vehicle", "Motorcycle"));
    rie.addRule(rule);

    rule=new Rule("SportsCar");
    rule.addAntecedent(new EqualsClause("vehicleType", "automobile"));
    rule.addAntecedent(new EqualsClause("size", "medium"));
    rule.addAntecedent(new EqualsClause("num_doors", "2"));
    rule.setConsequent(new EqualsClause("vehicle", "Sports_Car"));
    rie.addRule(rule);

    rule=new Rule("Sedan");
    rule.addAntecedent(new EqualsClause("vehicleType", "automobile"));
    rule.addAntecedent(new EqualsClause("size", "medium"));
    rule.addAntecedent(new EqualsClause("num_doors", "4"));
    rule.setConsequent(new EqualsClause("vehicle", "Sedan"));
    rie.addRule(rule);

    rule=new Rule("MiniVan");
    rule.addAntecedent(new EqualsClause("vehicleType", "automobile"));
    rule.addAntecedent(new EqualsClause("size", "medium"));
    rule.addAntecedent(new EqualsClause("num_doors", "3"));
    rule.setConsequent(new EqualsClause("vehicle", "MiniVan"));
    rie.addRule(rule);

    rule=new Rule("SUV");
    rule.addAntecedent(new EqualsClause("vehicleType", "automobile"));
    rule.addAntecedent(new EqualsClause("size", "large"));
    rule.addAntecedent(new EqualsClause("num_doors", "4"));
    rule.setConsequent(new EqualsClause("vehicle", "SUV"));
    rie.addRule(rule);

    rule=new Rule("Cycle");
    rule.addAntecedent(new LessClause("num_wheels", "4"));
    rule.setConsequent(new EqualsClause("vehicleType", "cycle"));
    rie.addRule(rule);

    rule=new Rule("Automobile");
    rule.addAntecedent(new EqualsClause("num_wheels", "4"));
    rule.addAntecedent(new EqualsClause("motor", "yes"));
    rule.setConsequent(new EqualsClause("vehicleType", "automobile"));
    rie.addRule(rule);

    return rie;
}

Infer more facts using forward chaining

public void testForwardChain()
{
    RuleInferenceEngine rie=getInferenceEngine();
    rie.addFact(new EqualsClause("num_wheels", "4"));
    rie.addFact(new EqualsClause("motor", "yes"));
    rie.addFact(new EqualsClause("num_doors", "3"));
    rie.addFact(new EqualsClause("size", "medium"));

    System.out.println("before inference");
    System.out.println(rie.getFacts());
    System.out.println();

    rie.infer(); //forward chain

    System.out.println("after inference");
    System.out.println(rie.getFacts());
    System.out.println();
}

Search for answer to a question using backward chaining

public void testBackwardChain()
{
    RuleInferenceEngine rie=getInferenceEngine();
    rie.addFact(new EqualsClause("num_wheels", "4"));
    rie.addFact(new EqualsClause("motor", "yes"));
    rie.addFact(new EqualsClause("num_doors", "3"));
    rie.addFact(new EqualsClause("size", "medium"));

    System.out.println("Infer: vehicle");

    Vector<Clause> unproved_conditions= new Vector<>();

    Clause conclusion=rie.infer("vehicle", unproved_conditions);

    System.out.println("Conclusion: "+conclusion);
}

Ask more questions when no sufficient facts are present

public void demoBackwardChainWithNullMemory()
{
    RuleInferenceEngine rie=getInferenceEngine();

    System.out.println("Infer with All Facts Cleared:");
    rie.clearFacts();

    Vector<Clause> unproved_conditions= new Vector<>();

    Clause conclusion=null;
    while(conclusion==null)
    {
        conclusion=rie.infer("vehicle", unproved_conditions);
        if(conclusion==null)
        {
            if(unproved_conditions.size()==0)
            {
                break;
            }
            Clause c=unproved_conditions.get(0);
            System.out.println("ask: "+c+"?");
            unproved_conditions.clear();
            String value=showInputDialog("What is "+c.getVariable()+"?");
            rie.addFact(new EqualsClause(c.getVariable(), value));
        }
    }

    System.out.println("Conclusion: "+conclusion);
    System.out.println("Memory: ");
    System.out.println(rie.getFacts());
}

private String showInputDialog(String question) {
    Scanner scanner = new Scanner(System.in);
    System.out.print(question + " ");
    return scanner.next();
}

Running rule engine using rules defined in a Javascript

Below is an example of a rules file written in Javascript (vehicle-rules.js)

expert.newRule("Bicycle")
    .ifEquals("vehicleType", "cycle")
    .andEquals("num_wheels", 2)
    .andEquals("motor", "no")
    .thenEquals("vehicle", "Bicycle")
    .build();

expert.newRule("Tricycle")
    .ifEquals("vehicleType", "cycle")
    .andEquals("num_wheels", 3)
    .andEquals("motor", "no")
    .thenEquals("vehicle", "Tricycle")
    .build();

expert.newRule("Motorcycle")
    .ifEquals("vehicleType", "cycle")
    .andEquals("num_wheels", 2)
    .andEquals("motor", "yes")
    .thenEquals("vehicle", "Motorcycle")
    .build();

expert.newRule("SportsCar")
    .ifEquals("vehicleType", "automobile")
    .andEquals("size", "medium")
    .andEquals("num_doors", 2)
    .thenEquals("vehicle", "Sports_Car")
    .build();

expert.newRule("Sedan")
    .ifEquals("vehicleType", "automobile")
    .andEquals("size", "medium")
    .andEquals("num_doors", 4)
    .thenEquals("vehicle", "Sedan")
    .build();

expert.newRule("MiniVan")
    .ifEquals("vehicleType", "automobile")
    .andEquals("size", "medium")
    .andEquals("num_doors", 3)
    .thenEquals("vehicle", "MiniVan")
    .build();

expert.newRule("SUV")
    .ifEquals("vehicleType", "automobile")
    .andEquals("size", "large")
    .andEquals("num_doors", 4)
    .thenEquals("vehicle", "SUV")
    .build();

expert.newRule("Cycle")
    .ifLess("num_wheels", 4)
    .thenEquals("vehicleType", "cycle")
    .build();

expert.newRule("Automobile")
    .ifEquals("num_wheels", 4)
    .andEquals("motor", "yes")
    .thenEquals("vehicleType", "automobile")
    .build();

The rule engine can then load these rules into its shell and run:

JSRuleInferenceEngine engine = new JSRuleInferenceEngine();
String jsContent = readToEnd("/vehicle-rules.js");
engine.loadString(jsContent);
engine.buildRules();

engine.clearFacts();

engine.addFact("num_wheels", "4");
engine.addFact("motor", "yes");
engine.addFact("num_doors", "3");
engine.addFact("size", "medium");



System.out.println("before inference");
System.out.println(engine.getKnowledgeBase());
System.out.println();



engine.infer(); //forward chain

System.out.println("after inference");
System.out.println(engine.getKnowledgeBase());
System.out.println();