AIwiki

Read:

• Reasoning on the Web
• OWL vs. RDF, DL, DL rules, DLP
• Rules on the Semantic Web

Reasoning on the Web - summary:

• queries
  • SPARQL (see RDF lab)
• DL reasoning
  • more powerful than querying, DL reasoning tasks
  • example reasoner: http://clarkparsia.com/pellet/, demo: http://pellet.owldl.com/owlsight/
• rules over ontologies
  • RuleML, RIF - rule interchange formats
  • SWRL Semantic Web Rule Language based on OWL
• OWL 2 RL - new version of OWL, additional constructs for expressing rules in the ontology

• Analyze the reasoning examples in OWL DL
• Use the Manchester Pizza Finder to see the results of implementing your ontology from the previous lab
• Browse sample ontologies with OwlSight (using Pellet as a DL reasoner)
  • Find all inconsistencies of the example ontologies.
  • Read the explanations of the inconsistencies - how could one repair the ontologies?

In this exercise we will use Protege editor with the SWRLTab plugin.

Depending on the configurations, Pellet/Jess may be used to execute SWRL rules.

1. Read about SWRLTab in Protege
2. Download family ontology with rules. Open it in Protege, anayze the rules. Enable JessTab (Project→Configure) and run the rules iteratively. Check the inferred axioms.
3. Open Protege editor and enable SWRL Tab (Project→Configure in Protege 3.4.4)
4. Implement an example expert system - an expert system for recognizing animals:
   1. define Classes used in the system (look for isa predicate): Mammal, Bird, Carnivore, etc.
   2. define characteristics of the animals (There Is More Than One Way To Do It!)
      • as Properties, e.g.: has
      • as Classes, e.g. Swims
   3. Make assertions about individuals, e.g. has(dennis,hair). parent(dennis,diana). Swims(denzel).
   4. Define rules of the system
   5. Express the initial data as appropriate assertions
   6. Run Pellet/Jess to execute the rules. Observe the inferred facts. Execute the rule engine iteratively, till no more information can be concluded.
   7. extra task: add more rules to the system, use built-in SWRL predicates.

rule id1: 
  [1: has(X,hair)]
  ==>
  [assert(isa(X,mammal)),
   retract(all)].

has(?x,„hair”) → Mammal(?x)

rule id4: 
  [1: flies(X),
   2: lays_eggs(X)]
  ==>
  [assert(isa(X,bird)),
   retract(all)].

Flies(?x) ∧ Lays_eggs(?x) → Bird(?x)

initial_data([
	has(dennis,hair),
	has(dennis,hoofs),
	has(dennis,black_stripes),
	parent(dennis,diana)
	]).

Read: OWL 2 and Rules slides

Control questions:

• How to express rules using owl:subClass relationships?
• what is a property chain axiom?

If Protege 4.1. is available:

1. Create a new ontology in Protege
2. Model the family ontology with rules using OWL 2 constructs
3. extra task: if there is enough time, model the animals expert system in OWL 2 RL

Further reading:

• Logics for the Semantic Web - FOL, DL, LP & F-logic

Test cases for OWL2 (incl.OWL2RL):

• http://www.w3.org/2007/OWL/wiki/Test_Suite_Status