OWL-XML-Summer-School-09

Ontologies, OWL and Prot égé Duncan Hull The University of Manchester, UK http://www. manchester .ac. uk Semantic Technologies Tuesday 22nd September 2009

Learning Objectives
Understand some of the Web Ontology Language (OWL) and it’s explicit semantics
Learn some of the principles of modelling using Description Logic ontologies and reasoning
Gain hands-on introductory experience with ontology development using Protégé-OWL tools
Learn how to take advantage of inferencing capabilites to build robust, reusable models
Review where OWL fits in with related technology and why you might want to use it This tutorial normally takes 1-2 days (10 exercises), we can only scratch the surface of OWL and Protégé in a 90 minute session (~2 exercises) http://bit.ly/owl-tutorial 11th/12th November, MAN
Short summary of the class

Course Contents
Why would you want to use OWL?
- The Web Ontology Language
What is an OWL ontology?
Where does OWL fit with related technology?
- W3C standards: XML, RDF and SPARQL
- Relational Databases
- Linked Data
Who is responsible for OWL?
When was OWL created?
How can you use OWL?
- Tutorial Session
Minor modifications to slides: http://www.slideshare.net/dullhunk

Where I’m coming from…
Trained as a Biologist
Scientific, technical publishing and software engineering
PhD Computer Science (OWL+Web Services) 2007
Now integrating and mining public biochemical data on the web
Tamiflu ChEBI:7799 http://www.sbml.org It’s all John’s fault!

Why would you want to use OWL?

Why?
Your data is important and you’re prepared to invest resources to precisely define the meaning in a way that computers can “understand” and infer additional information
Image via http://www.flickr.com/photos/dullhunk/639163558/

Why would you bother doing that?
1. Semantic Integration of Big Data
“ The Web is Agreement”
2. Better Search and Querying
“ Google is great but…”
3. Artificial Intelligence (A.I.)
“ A more knowledgeable web…”
4. Some examples…
5. Standardisation

Why? No. 1 Semantic Integration
Big Data: Lots of scenarios involve integrating data from multiple different sources:
In some cases, data integration is easier and quicker where semantics are agreed in advance (rather than cleaning it up afterwards)
CC-image via http://en.wikipedia.org/wiki/File:Datawarehouse.png
stuff

Why? The Web is Agreement
http://thewebisagreement.com/
OWL ontologies can be used to express agreement about the meaning of data on the Web
Between: Human-human, human-machine, and machine-machine
Paul Downey

Why? No. 2 Better search
is great but search engines could be much better , for example:
Complex queries involving background knowledge:
“ Find information about animals that use sonar but are neither bats or dolphins” (answer: barn owl)
Finding and using web services:
“ Book me a holiday next weekend somewhere warm, not too far away and where they speak french or english”
Usually impossible to do using the syntactic web search, need semantic search Image via http://en.wikipedia.org/wiki/File:Tyto_alba_close_up.jpg

Why? No. 3 A.I.
Artificial Intelligence was/is a key
motivation behind the semantic web
E.g. “A new form of Web content that is meaningful to computers will unleash a revolution of new possibilities” - Tim Berners-Lee et al , Scientific American, 2001

Why? no. 3: A.I.
2001: A Semantic Odyssey?
Realising the complete vision is (probably) too hard for now but we can make a start…
People have already started to build semantic web s (plural) rather than a monolithic semantic web, lets have a look at some of them…
HAL 9000 Semantic Web? I'm sorry, Dave. I'm afraid I can't do that.

Why? no. 4: Some examples:
Large biomedical terminologies using OWL
National Cancer Institute Thesaurus http://cancer.gov The NCI Thesaurus provides definitions, synonyms, and other information on nearly 10,000 cancers and related diseases
Contains 50,000 concepts managed by up to 20 people, provides terminology for applications like the cancer image database.

Why? no. 4. Some examples
“ SNOMED CT® is a clinical terminology - the Systematised Nomenclature of Medicine Clinical Terms. It is a common computerised language that will be used by all computers in the NHS to facilitate communications between healthcare professionals in clear and unambiguous terms.”
373,731 classes and over 1 million terms
NHS version extended to 542,380 classes
Large ontology classified in < 4 hours
Reasoner finds inconsistencies: e.g. 180 missing subclasses
Periocular_dermatitis subClassOf Disease_of_face

Why? 4. More examples
Pharmaceutical, biotechnology, drug discovery etc…
ChEBI: Ch emical E ntities of B iological I nterest
http://www.ebi.ac.uk/chebi
“ freely available dictionary of ‘small’ chemical compounds” (e.g. many drugs) using OWL
Currently contains ~500,000 small molecules, OWL is used to automate curation of the database and check quality

Why no.5 : Standardisation
There are plenty of different ontology languages:
… OWL is the only one that is a W3C standard…
Large and active community of developers and users around the world
Choice of tools to handle OWL
Interoperability
etc

Why? Summary
Biomedical applications
Healthcare and Life Sciences
Lots of terminology
Big data
Gigabytes / Terabytes of data
Manual curation not possible
Scientific applications (W3C HCLSIG)
Precision and Accuracy are important
http://www.w3.org/2001/sw/hcls/
“ Biology is just naming things”

“An ontology is a formal representation of a set of concepts within a domain and the relationships between those concepts. It is used to reason about the properties of that domain.”
http://en.wikipedia.org/wiki/Ontology_(information_science)

What? Pizza ontology
We’re going to use Pizzas in this tutorial…
We could use more realistic examples but they require a specialist knowledge of:
Biochemistry
Cancer
Medicine
etc
… Whereas we are all “experts” on Pizza, Pizzas are the “Hello world” of ontologies
http://www.co-ode.org/ontologies/pizza/
Pizza from http://www.flickr.com/photos/roadsidepictures/1544645159/

Pizza Margherita Pizza Vegetarian Pizza Spicy Beef Pizza What? A simple pizza ontology hasTopping (Object property) hasBase (Object property) subclassOf Pizza (A class in asserted hierarchy) Pizza Topping Vegetable topping Tomato topping Mozzarella topping Cheese topping Pizza_base Deep dish base Regular base

What? Object properties
Things you can say about properties: some , only , min , max and exactly
some means at least one of the toppings is a CheeseTopping
only means all of the toppings are CheeseTopping
min , max and exactly are self-explanatory

What? Object properties
More things you can say about properties
Symmetric e.g. touches (or spouse )
PizzaTopping touches PizzaBase
implies
PizzaBase touches PizzaTopping
Transitive e.g. subClassOf
Pizza subClassOf Food
CheeseyPizza subClassOf Pizza
implies
CheeseyPizza subClassOf Food
These are important for reasoning

What? Structure of ontology
Classes and properties: Terminology (TBox) Instances: Assertions (ABox)
TBox is a similar to a database schema e.g.
Pizza hasBase PizzaBase
VegetarianPizza hasTopping Vegetables
e.t.c.
Abox is similar to data (instances) in a database
ThisPizza is-an-instance-of CajunPizza
America is-an-instance-of Country
Fred is-an-instance-of DogLover
ABox + TBox combined called a “knowledgebase”

What? Logic and Reasoning
A key feature of OWL is reasoning (aka classification) , with a Description Logic (DL) reasoner (a bit like a source code compiler).
There are four basic tasks a reasoner can perform:
Subsumption : check that knowledge is correct
Equivalence : check for minimal redundancy
Consistency : check for contradictions
Instantiation : is a an instance of b ?
The reasoner infers new information from your asserted class hierarchy and builds a new inferred class hierarchy based on your definitions
Automates classification that might otherwise be done manually

What? OWL Subsumption
Check knowledge is “correct”
E.g. Fiorentina should be a subclass of VegetarianPizza?
If inferred hierarchy is inconsistent with intuition then this indicates an error
(in your model)

What? OWL equivalence
Similar to subsumption, a reasoner will tell you when two classes are equivalent e.g.
BoringPizza is equivalent to a MargheritaPizza
PizzaTopping is equivalent to PizzaBase ?

What? OWL Consistency
Consistency: check that no contradictory statements have been made: in Protégé these are highlighted in red
Cheese and Vegetable are disjoint classes (can’t be both)
CheeseyVegetable is a subclass of Cheese
CheeseyVegetable is a subclass of Vegetable

What? OWL instantiation
Check for instances of a class
E.g. Show me all the instances of CheeseyPizza
Important for querying (not covered here)

Where does OWL fit with related technology?

Where does OWL fit?
“An ontology is a formal representation of a set of concepts within a domain and the relationships between those concepts. It is used to reason about the properties of that domain.”
http://en.wikipedia.org/wiki/Ontology_(information_science)
Sounds a little bit like:
RDF and RDF Schema
Relational Databases
XML Schema
etc?
also Linked data

Where does OWL fit?
The Semantic Web: Will it all end in tiers?
Unicode + URIs + namespaces RDF/XML OWL/XML e.t.c. OWL 2.0 explicit semantics syntaxes OWL builds on top of standards you already know or have just learned about at the XML summer school SPARQL-DL To be done SPARQL Previous tutorial

Where? Relational databases
Some key differences between OWL and Relational Databases (DBMS)
Open World Semantics
Rejecting updates
Use of schema to answer queries
There are more differences, see Reference [6] “Ontologies and the semantic web” at the end for more details

Where? Open World
Open World Assumption: missing information is treated as unknown rather than false
c.f. Databases which make closed world assumption
In a social networking website, missing information (who all your friends are) is often treated as false: e.g. “ You have NO friends ” (loser!) This is a subtle but important distinction On http://network.nature.com/people/duncan

Where? Rejecting updates
Unlike DBMS, ontology tools typically don't reject updates that result in the ontology becoming inconsistent, they just warn.
You’ll see this in the exercises…

Where? Query answering
In OWL, the schema plays a much more important role and is actively considered at query time (but discarded with DBMS) - this makes it possible to answer conceptual queries e.g.
Pizza from http://www.flickr.com/photos/roadsidepictures/1544645159/ Is any Pizza that hasTopping Cheese necessarily a CheeseyPizza?

Where? OWL and Linked Data
Use URIs to identify things that you expose to the Web as resources. YES, everything important has a URI
Use HTTP URIs so that people can locate and look up (dereference) these things. YES, but don’t have to
Provide useful information about the resource when its URI is dereferenced. YES, but don’t have to
Include links to other, related URIs in the exposed data as a means of improving information discovery on the Web. YES, but again this is optional
Returning to the linked data session… http://www.co-ode.org/ontologies/pizza/ for an example of owl and linked data

Who is responsible for OWL?

Who?
OWL is managed by a Working Group at the W3C
http://www.w3.org/2007/OWL/
A large group of people chaired by :
http://web.comlab.ox.ac.uk/ian.horrocks/ http://sciencecommons.org/about/whoweare/ruttenberg/ Ian Horrocks, The University of Oxford Alan Ruttenberg, Science Commons

Who?
Every year developers and users of OWL gather at OWLED ( OWL E xperiences and D irections)
http://www.webont.org/owled/
5th International Workshop on 23-24th October 2008, Chantilly, Virginia, USA
Co-located with the 8th International Semantic Web Conference (ISWC) 25-29th October http://iswc2009.semanticweb.org/ Washington, DC, USA.

When was OWL created?

When?
OWL 1.0 a recommendation in 2004
http://www.w3.org/2004/OWL/
http://www.w3.org/TR/owl-semantics/
OWL 2.0 a candidate recommendation in 2009
http://www.w3.org/TR/owl2-profiles/
See “OWL 2.0: The next step for OWL” in the references at the end…
Compare that to XML 1.0 which was a recommendation in 1998…

When?
But ontologies generally are much older than that…
CC picture from http://en.wikipedia.org/wiki/File:Sanzio_01_Plato_Aristotle.jpg A Aristotle Οντολογία
Linguistics
Natural Language Processing (NLP)
Philosophy
Data mining
Text mining

How can you use OWL?

How? Protégé
Protégé is a free, Open Source ontology editor
http://protege.stanford.edu/
http: //protege . stanford .edu/download/protege/4.0/installanywhere/
Protégé research & development has been led by
Professor Mark Musen
Stanford University, USA
Professor Alan Rector
University of Manchester, UK
Protégé supports latest version of OWL (OWL 2.0) and uses the OWL-API http://owlapi.sourceforge.net/
Mark Musen Alan Rector

How? Hands-on tutorial
See tutorial slides at the end
Don’t worry if you can’t complete all the exercises, there is one exercise too many, just in case.

Acknowledgements
John Chelsom and Lauren Wood
Information Management Group (IMG) and Bio-Health Informatics Group (BHIG) at The University of Manchester: Alan Rector, Matthew Horridge, Simon Jupp, Nick Drummond, Robert Stevens, Holger Knublauch, Georgina Moulton, Chris Wroe, Ulrike Sattler, Ian Horrocks, Bijan Parsia, Sean Bechhofer, Carole Goble and many others
Currently funded by www.bbsrc.ac.uk as part of REFINE project www.nactem.ac.uk/refine devised by Douglas Kell and Sophia Ananiadou
substantial parts of this tutorial and slides have been developed by the http://www.co-ode.org/ project with funding from www.jisc.ac.uk

Any questions? Thank you for your attention

How? Protégé tutorial ex. 2
Start Protégé
Click on “Open OWL Ontology”
Open the exercise 2 ontology: select “pizza-ex2.owl” from the exercise folder (exercise 1 of building this ontology has been done for you to save time)
Explore the “asserted class hierarchy” by clicking on the classes in the “classes” tab
Add some new subclasses by selecting MeatTopping and then pressing “Add Subclass” button (top left button in asserted classes hierarchy)

How? Protégé tutorial, ex. 2 cont.
Note that the MeatyVegetableTopping has been asserted to be a subclass of both Meat and Vegetable (see “Superclasses” in the “Description” pane on right hand side. Is this inconsistent?)
Click on the “Inferred class hierarchy” and note that it should be empty (apart from a single class called “Thing”)
On the “Reasoner” menu, select a reasoner (there are different reasoners available but “FaCT++” is easiest to use for this exercise)
Now select “Classify” from the same menu, this will run the reasoner. What is the result?
Save the result, (note the different available syntaxes for saving ontologies)

How? Protégé tutorial ex. 2 cont.
To make sure toppings can not be both meat and vegetable at the same time you need to add disjoint axioms to explicitly state the disjunction.
Select one of your top level concepts (e.g. Pizza) and press ctrl-J (windows) or cmd-j (mac) to make Pizza disjoint from all its sibling classes
Note that the “Description” pane lists all classes Pizza is now disjoint with (e.g PizzaBase etc)
Repeat this for the MeatTopping level of the ontology
Run the reasoner, is MeatyVegetableTopping now inconsistent as expected?

How? Protégé tutorial ex. 4
Exercise 3 has been skipped, close your current ontology and the solution “pizza-ex3.owl” to start ex. 4
Exercise 4 is included here for keen students (and anyone wanting to do some homework)
In order to describe our classes more fully we need properties which relate members of a class. We can then add restrictions on the class to state how the properties are used.
At this stage we are creating Primitive Classes , which only have Necessary Conditions . These are conditions that must be satisfied by all members of this class
Select the “Object Properties” called hasTopping (this is a relation between two classes).
Back in the classes tab, create a new subClass of Pizza called NamedPizza
Create a new subClass of NamedPizza called MargheritaPizza

How? Protégé ex. 4 continued
Create restrictions on MargheritaPizza: In the “Description” pane under “Superclasses” click on the “+” button to add a restriction
Type “hasTopping some MozzarellaTopping” (This says that it is necessary condition for a MargheritaPizza, to have at least one Topping that is a MozzarellaTopping)
Repeat this process to state that this kind of pizza also “hasTopping some TomatoTopping”
Run the reasoner to check for consistency

References
1. Protégé is a free, open source ontology editor and knowledge-base framework that is available from http://protege.standord.edu The version you have been using in this tutorial is Protégé 4.x
2. The CO-ODE project http://www.co-ode.org has lots more useful material on ontologies. For example, a complete finished version of the Pizza and other ontologies are available from http://www.co-ode.org/ontologies/
3. Matthew Horridge (2004) Protégé OWL Tutorial. This is a comprehensive guide to OWL, more complete than this tutorial and available from http://owl.cs.manchester.ac.uk/tutorials/protegeowltutorial/ see some more examples at http://owl.cs.manchester.ac.uk/2009/07/sssw and software http://owl.cs.manchester.ac.uk
4. Alan Rector, Nick Drummond, Matthew Horridge, Jeremy Rogers, Holger Knublauch, Robert Stevens, Hai Wang, Chris Wroe (2004) OWL Pizzas: Practical Experience of Teaching OWL-DL: Common Errors and Common Patterns In Proc. of European Conference on Knowledge Acquistion (EKAW'04), Vol. 3257 (2004), pp. 63-81. http://www.co-ode.org/resources/papers/ekaw2004.pdf gives an overview of common errors and pitfalls (with solutions) to building ontologies in OWL using pizzas as an example
5. Ian Horrocks (2003) From SHIQ and RDF to OWL: the making of a Web Ontology Language Journal of Web Semantics: Science, Services and Agents on the World Wide Web, Vol. 1, No. 1. (December 2003), pp. 7-26. (this paper gives a readable overview of the relationship between RDF and OWL with some history on the development and integration of the two languages) A free version of this paper is available from http://www.comlab.ox.ac.uk/people/ian.horrocks/Publications/download/2003/HoPH03a.pdf
Ian Horrocks (2008) Ontologies and the semantic web. Commun. ACM, Vol. 51, No. 12, pp. 58-67. http://www.comlab.ox.ac.uk/people/ian.horrocks/Publications/download/2008/Horr08a.pdf gives an nice overview of the differences between OWL and relational databases
7. These and other papers relating to OWL, Ontologies and Protégé are available in citeulike tagged as “xml summer school” at http://www.citeulike.org/tag/xml-summer-school

OWL-XML-Summer-School-09

by Duncan Hull on Sep 22, 2009

Accessibility

Categories

Tags

Upload Details

Usage Rights

1 Embed 16

Statistics

OWL-XML-Summer-School-09 — Presentation Transcript