eXtreme Design
– “rapid prototyping” of
ontologies
Eva Blomqvist
eva.blomqvist@liu.se
Department of Computer and Information Science (IDA)
Linköpings universitet
Sweden
Slides partly by Valentina Presutti, STLab, ISTC-CNR, Italy
September 24, 2012
1
Example: METHONTOLOGY (~1997)
n
Waterfall-like process consisting of (overlapping) phases
1.
2.
3.
4.
5.
6.
7.
Specification – document requirements, scope, level of
formality etc.
Knowledge Acquisition – gathering and studying sources
of information
Conceptualization – structure the terminology identified in
1, going from glossary to logical formulas
Integration – find and select other ontologies to reuse
Implementation – represent in formal language using tool
Evaluation – verification and validation
Documentation
Department of Computer and Information Science (IDA)
Linköpings universitet, Sweden
September 24, 2012
eXtreme Design (XD)
– Method and tool support
n
eXtreme Design (XD)
q
n
a method for developing ontologies with Content Patterns
XD tools
a tool that supports the XD method
q  released as both an Eclipse plugin and a NeOn Toolkit
plugin
q  Can be used with the NeOn toolkit version 2.3.2 and older
http://neon-toolkit.org/wiki/Download/2.3.2
q
Department of Computer and Information Science (IDA)
Linköpings universitet, Sweden
September 24, 2012
eXtreme Design – General idea
n
n
XD is a general approach to ontology engineering
Local Use Case (LUC)
q
n
Generic Use Case (GUC)
q
n
n
represents the current modeling issue
represents a generic problem that is solved by the associated ODP
GUC and LUC are represented in a compatible comparable
way – Competency Questions?
What ODP to reuse?
q
q
The one where LUC matches GUC
Note: GUC are often more abstract than LUC
Department of Computer and Information Science (IDA)
Linköpings universitet, Sweden
September 24, 2012
What kinds of “requirements” are used to
describe use cases?
n
n
Viewing an ontology as a black box…
what should that box provide?
“Functional requirements”
q
q
q
q
n
Query results?
Inferences?
Error checking?
…
Internal structure,
and content
“Non-functional requirements”
q
q
q
q
q
Coverage
Efficiency
Documentation
Changeability – extendibility
…
Department of Computer and Information Science (IDA)
Linköpings universitet, Sweden
September 24, 2012
5
Overall structure, acceptance
è Guidelines and rules for
development
Requirements Engineering
– Competency Questions
n
Competency Questions (CQ) = Natural language questions
that ask for information the ontology should be able to
provide to a user (or system)
q
q
n
C.f. functional requirements for software
Related to software requirements – “input” and “output” of the
“ontology component” (including query engine, reasoner…)
Different kinds
q
Simple lookup queries
n
q
Who are the participants of a certain course?
Expressing inferences or constraints
n
n
Given that people may have children, is a specific person a
grandparent or not?
Is a person married to two people valid according to Swedish law?
Department of Computer and Information Science (IDA)
Linköpings universitet, Sweden
September 24, 2012
6
Requirements Engineering
– Competency Questions (cont.)
n
To clarify complex CQs we use
q
q
n
A contextual statement expresses an axiom that needs to
hold in the ontology, in natural language
q
q
q
n
Contextual statements
Inference (reasoning) requirements
Every course has at least one participant
A grandparent is someone who has a child who in turn also
has a child
In Sweden you can only marry one person
Reasoning requirements specify the input and output
data for a reasoning task
q
We would like to be able to query directly for all the
grandparents – classification based on the axiom above
Department of Computer and Information Science (IDA)
Linköpings universitet, Sweden
September 24, 2012
7
Reuse
n
Ontology Design Patterns
q
q
q
C.f. software design patterns
Different kinds – for different purposes
Content ODPs – solves modelling problems in a domain
n
n
n
Available as components
Ontologydesignpatterns.org
Existing ontologies
q
q
You have already used FOAF for other task
Search engines for finding ontologies, e.g.
http://kmi-web05.open.ac.uk/WatsonWUI/
Department of Computer and Information Science (IDA)
Linköpings universitet, Sweden
September 24, 2012
8
Matching GUC and LUC
n
A LUC can be completely or partly described exactly in
terms of the GUC
q
q
n
A LUC is a more specific case of the GUC
q
q
n
Guc: Performing in a concert
Luc: John Coltrane performed in a concert in Japan in 1966
Guc: Participating in an event
Luc: Mary attended a scientific conference
A LUC can be described in terms of part of the GUC
q
q
Guc: Participating in an event held in a certain place at a
certain time
Luc: Mary attended a conference in Italy
Department of Computer and Information Science (IDA)
Linköpings universitet, Sweden
September 24, 2012
Why the name “XD”?
n
n
n
Inspired by XP J with focus on design
An agile methodology for web ontology design
It is part of the NeOn methodology
Department of Computer and Information Science (IDA)
Linköpings universitet, Sweden
September 24, 2012
XD principles
n
n
n
n
n
n
n
Customer involvement and feedback
Customer stories to derive CQs (+ contextual statements,
reasoning requirements)
CP reuse and modular design (ontology networks)
Collaboration and integration
Task-oriented design
Test-driven design
Pair design
Department of Computer and Information Science (IDA)
Linköpings universitet, Sweden
September 24, 2012
XD
Iteration
Project((
idea(
Project(ini2a2on((
and(scoping(
Iden2fying(CP((
catalogues(
Design(team(
CP((
catalogues(
Seman2c(
Web(
Design(team(
Collec2ng((
requirement(stories(
Stories(
Customer(
Design(team(
Selec2ng(
story(
Design(pair(
Elici2ng(requirements(
and(construc2ng(
module(s)((
from(CPs(
No(
Releasing((
module(s)(
All(stories(
covered?(
Integra2ng(par2al(
solu2ons,(evalua2ng((
and(revising(
Integra2on(team(
Department of Computer and Information Science (IDA)
Linköpings universitet, Sweden
September 24, 2012
Releasing((
new(version(of(
Ontology(Network(
Ontology((
Network(
Customer(
XD
Iteration
Project((
idea(
Project(ini2a2on((
and(scoping(
Iden2fying(CP((
catalogues(
Design(team(
CP((
catalogues(
Seman2c(
Web(
Design(team(
Collec2ng((
requirement(stories(
Select'
story'
Stories(
Customer(
Design(team(
Elici$ng''
requirements'
Selec2ng(
story(
Select'
set'
Matching'and''
selec$ng'pa8erns'
Design(pair(
Elici2ng(requirements(
and(construc2ng(
module(s)((
from(CPs(
Reusing'and'
integra$ng'CPs'
No(
No'
Tes$ng''
module''
Releasing((
module(s)(
All'require=
ments'
covered?'
All(stories(
covered?(
No'
Releasing''
module(s)'
All'stories'
covered?'
Integra2ng(par2al(
solu2ons,(evalua2ng((
and(revising(
Integra2on(team(
Department of Computer and Information Science (IDA)
Linköpings universitet, Sweden
September 24, 2012
Releasing((
new(version(of(
Ontology(Network(
Ontology((
Network(
Customer(
XD – Method steps
n
Tasks 1-2: Project initiation and scoping,
collecting resources
q
q
q
q
n
Task 3 – Collect requirements stories
q
q
n
Essential to understand the context and task of the
ontology
Customer involvement – domain experts
Setup the project environment (collaboration
support) + the resources to be used
Agree on general “rules”, e.g. naming conventions
Example scenarios (cf. the stories of the exercise)
Should be short and “modular”
Select a story (each design pair!)
Department of Computer and Information Science (IDA)
Linköpings universitet, Sweden
September 24, 2012
XD – Method steps
n
Task 4.1 – Transform the story into CQs
q  Divide&Conquer strategy in the large
– each pair works on one story at the time
Stories are associated with priority values
n  Based also on design pair competencies
Derive requirements from the text
n
q
n
n
q
n
CQs, contextual statements and reasoning requirements
Check with customer representative & SW developers
Should correspond to actual queries/tasks that the user/
system need to pose/perform
Select a CQ (each pair iterates)
q
q
q
…or coherent set of CQs
Together with associated contextual statements & reasoning
requirements
Divide&Conquer strategy in the small – pair iterates over
CQs and creates module(s)
Department of Computer and Information Science (IDA)
Linköpings universitet, Sweden
September 24, 2012
XD – Method steps
n
Task 4.2 – Matching and selecting ODPs
q
q
q
q
q
q
q
q
How? Either only intellectually or with some
tool support e.g. XD Selector
Can I describe my local problem in terms of the
general problem of the ODP?
Does the ODP solve the same “design issue”?
Partial match – Is it worth the overhead?
Several ODPs needed – Composition of ODPs
In case there is no matching ODP – consider to
create one!
May exist several options
“Rule of thumb” – use the most (domain)
specific one applicable
Department of Computer and Information Science (IDA)
Linköpings universitet, Sweden
September 24, 2012
XD – Method steps
n
Task 4.3 – Reuse and integrate
selected Content ODPs
q
q
q
q
Specialize
Import
Extend
Integrate (compose)
Department of Computer and Information Science (IDA)
Linköpings universitet, Sweden
September 24, 2012
XD – Method steps
n  Task 4.3. Test and Fix – Unit tests
1.
SPARQL queries
Assume the following CQ:
What role did a certain person play in the production of a certain play during a
certain time period?
SELECT ?person ?role ?play ?startTime ?endTime
WHERE {
?rolePlaying a :PlayingSituation .
?rolePlaying :personPlayingRole ?person .
?rolePlaying :rolePlayed ?role .
?rolePlaying :playedInProduction ?production .
?production :productionOfPlay ?play
?rolePlaying :playedDuringTime ?timeInterval .
?timeInterval :hasStartDate ?startTime .
?timeInterval :hasEndDate ?endTime .
}
2.
3.
Producing inferences
“Stress testing”
Department of Computer and Information Science (IDA)
Linköpings universitet, Sweden
September 24, 2012
XD – Method steps
n
Task 5 – Release module
q
q
q
q
n
Make sure the module is commented and ready
Post the module so that it is accessible by the other pairs
Post any new reusable Content ODP developed
Taken over by integration pair
Task 6 – Integrate, test and fix (integration team)
q
Integrate with overall ontology so far
n
n
q
n
Alignment may be needed
Refactoring may be needed
Run all unit tests based on all included requirements
Task 7 – Release new version of the ontology
q
q
q
Distribute
Generate documentation
Check customer satisfaction, i.e. evaluation of overall
ontology including non-functional requirements
Department of Computer and Information Science (IDA)
Linköpings universitet, Sweden
September 24, 2012
XD Summary
n
n
n
n
XD is an agile method – start building small modules that
solve a few requirements, then add more, we don’t
decide on all the requirements at once
Testing is essential – by figuring out the test you figure
out how the model should work!
Collaboration is essential
Many problems are resolved in the integration phase –
alignments or refactoring?
q
n
Need for good overall design policies
You are about to experience the method!
Department of Computer and Information Science (IDA)
Linköpings universitet, Sweden
September 24, 2012
Exercise
n
n
n
Build an ontology collaboratively
From scenario, via requirements to implemented model
Follow the XD methodology
q
q
q
q
Derive requirements
Reuse
Model
Document
n
n
q
n
Try to finish one module at least by 2pm
Joint hand-in by all groups
q
q
n
Release and integrate
Second group who is finished – integration team!
q
n
In the module use RDFS-annotation properties and/or an annotation
schema like
http://www.ontologydesignpatterns.org/schemas/cpannotationschema.owl
In the wiki (request an account to be able to edit page):
http://ontologydesignpatterns.org/wiki/
Training:Semantic_Technologies_in_Practice_PhD_course_in_Link
%C3%B6ping_2012/XD_collaborative_OE
Each group sends what they have produced
Integration team sends integrated ontology to everybody
Short discussion at 15:50
Department of Computer and Information Science (IDA)
Linköpings universitet, Sweden
September 24, 2012
21
Department of Computer and Information Science (IDA)
Linköpings universitet, Sweden
September 24, 2012
22