1. The Semantic Web Week 15 Reasoning with (and Visualising) Ontologies Module Website: http://scom.hud.ac.uk/scomtlm/chs2533 Practical :Protégé-2000 WITH OWL: Work through Protégé tutorial up to page 73

2. Recap n Protégé-2000 is a Knowledge Acquisition Tool that helps one build up Ontologies in OWL. We can create hierarchies of is-a classes, properties, and necessary conditions for class membership. n The Pizza example is useful for learning about ontologies as it contains a wide range of structured and composite types.

3. Question.. n Protégé is a tool which helps people build ontologies in OWL – that is enter knowledge in a form that can be used in the Semantic Web. Question: Can non-computer experts use it?

4. CONTINUED: Developing ontologies in OWL: “Partial” and “Complete” Definitions n Partial = Necessary Condition n Complete = Necessary and Sufficient Condition Example: set {2,4,6,8,10,12} Necessary condition for membership: even number [but 14 is even but not a member..] Sufficient condition for membership: = 2 N, where 0 < N < 4 [but 12 is a member that is not = 2 N ] Necessary and Sufficient Condition: Even number between 2 and 12 inclusive.

5. OWL: “Partial” and “Complete” Definitions OWL abstract syntax: Class(a:Pizza partial restriction(a:hasBase someValuesFrom (a:PizzaBase))) FOL: Ax Pizza(x) => Ey hasBase(x,y) & PizzaBase(y) Pizzas PizzaBase Things that have at least one PizzaBase hasBase

6. Necessary and Sufficient Classes.. We need sufficient conditions - If we only have necessary conditions then we can never state for definite that an instance is a member of a class. Eg CheesyPizza(x)  Pizza(x) & Ey hasTopping(x,y) & CheeseTopping(y)

7. For-All Restrictions Vegetarian condition… Ay hasTopping(x,y) => (CheeseTopping(y) V VegetableTopping(y))..but also need an existential condition saying there exists at least one topping…

8. For-All Restriction gives Complete Defn: VegetarianPizza(x)  Pizza(x) & Ay hasTopping(x,y) => (CheeseTopping(y) V VegetableTopping(y)) NB there are other possible defns.. VegetarianPizza(x)  Pizza(x) & not (Ey hasTopping(x,y)&MeatyTopping(y)) & not (Ey hasTopping(x,y)&FishyTopping(y))

9. Reasoning- Subsumption n Used to RE-CLASSIFY an asserted hierarchy. The re-classification is called the inferred model. n Used for Consistency checking – can every class defined have some instances? Eg IS-A DISJOINT

12. Another Example: The Martians Application 1. GreenMartians C AntennaeMartians A Martian has antennae IF it is green. 2. A_has-child.Antennae C Friendly A Martian is friendly to humans IF all of its children have antennae. 3. E_has-parent.Green C Green A Martian is green IF at least one of its parents is green. Step 1: Define Martian class, and Green, Antennae, Friendly subclasses. Step 2: Create subclasses of Friendly and Green with necessary and sufficient conditions = A_has-child.Antennae and E_has-parent.Green respecively Step 3: Create has-parent and has-child properties Step 4: A necessary condition for a Martian is that it has at least one parent

13. The Martians Application- Asserted Model

14. The Martians Application Step 5: Run the classifier a) Use OWLViz – the asserted model is the facts that have been asserted. b) The inferred model should show that all green Martians friendly (the hierarchy has been changed)

15. The Martians Application – Inferred Model

16. Practical Work – 1. Continue with the protégé – owl tutorial. You should be up to page 49 now. This week do pages 50 – 73. Under “Project” configure Protégé with “OWLViz” 2. Put Martians into Protégé, and do reasoning using Protege