(12) Semantic Web Technologies - Ontological Engineering
-
Upload
harald-sack -
Category
Education
-
view
1.882 -
download
3
Transcript of (12) Semantic Web Technologies - Ontological Engineering
Semantic Web Technologies
LectureDr. Harald Sack
Hasso-Plattner-Institut für IT Systems EngineeringUniversity of Potsdam
Winter Semester 2012/13
Lecture Blog: http://semweb2013.blogspot.com/This file is licensed under the Creative Commons Attribution-NonCommercial 3.0 (CC BY-NC 3.0)
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
Rules
&
the Se
mantic
Web
2
last lecture
Dienstag, 15. Januar 13
Semantic Web Technologies , Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
3 1. Introduction 2. Semantic Web - Basic Architecture
Languages of the Semantic Web - Part 1
3. Knowledge Representation and LogicsLanguages of the Semantic Web - Part 2
4. Applications in the ,Web of Data‘
Semantic Web Technologies Content
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
4
Ontolo
gical
Engine
ering
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
4. Applications in the Web of Data4.1.Ontological Engineering4.2.Linked Data Engineering 4.3.Semantic Search
Semantic Web Technologies Content
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
6
4.Ontology Engineering4.1.Ontologies revisited4.2.Methodologies of Ontology Design
4.2.1. In General
4.2.2. Method of Uschold and King4.2.3. Ontology 1014.2.4. Unified Process for ONtology (UPON)
4.2.5. Ontology Design Patterns
4.3.Ontology Learning4.4.Ontology Mapping and Ontology Merging
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
7
What is an Ontology?
„A theory of being, which tries to explain the being itself, by developing a system of universal categories and their intrinsic relationships...“
PhilosophyDienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
8
What is an Ontology?
"An ontology is an explicit, formal specification of a shared conceptualization. The term is borrowed from philosophy, where an Ontology is a systematic account of Existence. For AI systems, what ‘exists’ is that which can be represented.“
Computer ScienceDienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
9
Domain Ontology Task Ontology
Application Ontology
(acc. to Guarino,1998)
Top-Level Ontology(Upper Ontology,
Foundation Ontology)
What is an Ontology?
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität PotsdamTurmbau zu Babel, Pieter Brueghel, 1563
10
Ontologies and the Semantic Web
•Semantic Web is based on the Interoperability of Metadaten
•Among heterogeneous Metadata there is a Semantic Gap that can be bridged with the help of ontologies
•Problem of the Semantic Gap:
• different ontologies can be applied to represent identical knowledge.
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität PotsdamTurmbau zu Babel, Pieter Brueghel, 1563
11
The Semantic Gap - A Simple Example
•Let‘s model a world:
A
C
B
Initial State Final State
A
C
B
World
Modelling 2:Objectsblock Ablock Bblock C
Relationson(X,Y)clear(X)onTable(X)holding(X)handEmpty
Modelling 1:Objectsblock Ablock Bblock Ctable Thand H
Relationson(X,Y)clear(X)holding(X)handEmpty
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität PotsdamTurmbau zu Babel, Pieter Brueghel, 1563
12
Modelling and Ontologies
A
C
B
Initial State Final State
A
C
B
•behind the model there is an ontology
Modelling 1:Objectxblock Ablock Bblock Ctable Thand H
Relationson(X,Y)clear(X)holding(X)handEmpty
⊤
entity relation
table block hand binary unary
handEmptyclear
holding
on
table T hand A
block A
block B
block C
Axiom: on(X,Y) ⋀ on(Y,Z) → above(X,Z)
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
13
⊤
entity relation
block binary unary
handEmptyclear
holding
onblock A
block B
block C onTable
Modelling and Ontologies•behind the model there is an ontology
A
C
B
Initial State Final State
A
C
BModelling 2:Objectsblock Ablock Bblock C
Relationson(X,Y)clear(X)onTable(X)holding(X)handEmpty
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
14•Ontologies enable interoperability among metadata•Therefore, we need•Methods for efficient development of ontologies
(Ontology Design)•Methods for efficient comparison of ontologies
(Ontology Mapping)•Methods for efficient combination of ontologies
(Ontology Merging)
•There are automated methods to support Ontological Engineering:•Learning new ontologies from a given set of information resources
(Ontology Learning)•Populating existing ontologies with individuals from information
resources
Ontological Engineering
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
15
„The only method to be driven, the empiricism is annoying.“
-- Johann Wolfgang von Goethe, aus „Maxims and Reflections”
„Zur Methode wird nur der getrieben, dem die Empirie lästig wird.“
-- Johann Wolfgang von Goethe, aus „Maximen und Reflexionen”
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
16
4.Ontology Engineering4.1.Ontologies revisited4.2.Methodologies of Ontology Design
4.2.1. In General
4.2.2. Method of Uschold and King4.2.3. Ontology 1014.2.4. Unified Process for ONtology (UPON)
4.2.5. Ontology Design Patterns
4.3.Ontology Learning4.4.Ontology Mapping and Ontology Merging
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
17
Methodologies of Ontology Design
•A methodology of Ontology Design describes all activities necessary for the construction of an ontology.
•Why do we need a formal methodology?
•development of consistent ontologies
•efficient development of complex ontologies
•distributed development of ontologies
•We distinguish (acc. to Fernandez-Lopez et. al., 1997)
•Ontology management activities
•Ontology development oriented activities
•Ontology support activities
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
18
Ontology Management Activities
•Scheduling•Identification tasks/problems to solve
•Arrangement/Planning of tasks/problems to solve
•Identification of required resources (time, memory, etc…)
•
•Control•Guarantees correct execution of tasks/problems to solve
•
•Quality Assurance•Quality assurance of all produced resources during
development(Ontologies, Software, Documentation)
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
19
1.Pre-Development2.Development3.Post-Development
1.Pre-Development•Environment Study
•What is the designated software platform for the ontology?
•Which applications should use the ontology?
•Feasibility Study
•Can the ontology really be developed?
•Does it make sense to develop the ontology?
Ontology Development Oriented Activities
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
20
Ontology Development Oriented Activities
2.Development
•Specification•Why is the ontology developed, what is the benefit and
who are the end-users?
•Conceptualization•Structuring domain knowledge in a conceptual model
•Formalization•Formalize conceptual model in (semi-)computable
model
•Implementation•Construction of a computable model in an ontology
representation language
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
21
Ontology Development Oriented Activities
3.Post-Development
•Maintenance•Update and adjustment of the ontology (if necessary)
•Use / Reuse•Usage of the ontology within the designated applications
as well as in unplanned applications/ontologies
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
22
Ontology Support Activities
•Knowledge Acquisition•Gather knowledge from experts (Ontology Learning)
•Evaluation•Technical evaluation of the ontology in each step of the
development process
•Integration•Reuse of existing ontologies (Ontology Reuse)
•Merging•Construction of a new ontology from already existing ontologies
within a specific domain
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
23
Ontology Support Activities
•Alignment•Mapping rules for involved ontologies
•Documentation •Each step of the ontology development must be acurately
documented
•Configuration Management•Manages all versions of documentation and of the developed
ontology
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
24
Ontological Engineering
Management Development Oriented Support
scheduling
control
quality assurance
environment study feasibility study
conceptualizationspecification
formalization implementation
maintenance use / reuse
knowledge acquisition
evaluation
documentation merging
alignmentconfigurationmanagement
integration
acc. to Fernandez-Lopez et. al., 1997
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
25
Ontological Engineering
GREATONTOLOGY
SOmething
Development Process
acc. to http://geekandpoke.typepad.com/geekandpoke/2012/01/simply-explained-dp.html
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
26
4.Ontology Engineering4.1.Ontologies revisited4.2.Methodologies of Ontology Design
4.2.1. In General
4.2.2. Method of Uschold and King4.2.3. Ontology 1014.2.4. Unified Process for ONtology (UPON)
4.2.5. Ontology Design Patterns
4.3.Ontology Learning4.4.Ontology Mapping and Ontology Merging
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
27
Methode of Uschold and King
•Process based development
•1995/96, first proposal of a methodology for ontology development
•IBM, University of Edinburgh, Unilever,...•Development of an ,Enterprise Ontology‘
identifypurpose capture coding integrating documen-
tationevaluation
Building
1 2 3 4
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
28
Methode pf Uschold und King
Identify Purpose and Domain of Application•Why is the ontology needed?•Designated application?
•(use / reuse / share / used as part of KB / …)•Identify all terms relevant for the application
identifypurpose capture coding integrating documen-
tationevaluation
Building
1
Example: Travel Ontology•Development of a common ontology for the domain ,Travel‘ that should be used in travel offices•Ontology could also be used for other application areas, e.g. to develop a catalogue for hotels and transportation means•relevant terms e.g: locations, types of locations, accommodations, types of accommodations (hotel / motel / camping / …), railway, busses, subway,...
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
29
Methode of Uschold and Kingidentifypurpose capture coding integrating documen-
tationevaluation
Building
Ontologie Building• Ontology Capture
• Humans as domain experts, who possess the mandatory knowledge...
• ...are not neccessarely fully-trained logicians that are able to design an ontology
• Therefore, knowledge engineers often are appointed to support the domain experts
2
Identify key concepts (classes) and relationships (relations) of the domain under consideration
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
30
Methode of Uschold and Kingidentifypurpose capture coding integrating documen-
tationevaluation
Building
Ontologie Building•Ontology Capture•Identify key concepts (classes) and relationships (relations) of the domain under consideration and provide them as plain text
•Identification of ontology concepts•Bottom-up / Top-down / Middle-Out
2
Example: Travel Ontology• Transportation is a class. Each Transportation has a Starting Point• Bus is a class. Bus is a Transportation.• City Bus is a class. A City Bus is a Bus, whose Starting Point as well as its Destination and all Stopovers are located in the same City.
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
31
Methode of Uschold and Kingidentifypurpose capture coding integrating documen-
tationevaluation
Building
Ontologie Building•Ontology Capture
•Bottom-up identification of ontology concepts •Construction from ,special‘ to ,general‘ •Identification of concepts with the most clear
specification, then generalisation towards abstract concepts
2
Example: Travel Ontology• Transportation is conceptualized with a Bottom-up strategy
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
32
Methode of Uschold and Kingidentifypurpose capture coding integrating documen-
tationevaluation
Building
Ontologie Building•Ontology Capture
•Bottom-up identification of ontology concepts
2
LondonUnderground
LondonLocal Bus
LondonTaxi
ParisMetro
ParisLocal Bus
ParisTaxi
London Transportation Paris TransportationSubway City Bus Taxi
Transportationis-subClass-of
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
33
Methode of Uschold and Kingidentifypurpose capture coding integrating documen-
tationevaluation
Building
Ontologie Building•Ontology Capture
•Bottom-up identification of ontology concepts
•increased total cost•difficult to find common ground among related
concepts•increased risk of inconsistencies→ then revision is necessary (even more expensive)
2
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
34
Methode of Uschold and Kingidentifypurpose capture coding integrating documen-
tationevaluation
Building
Ontologie Building•Ontology Capture
•Top-down identification of ontology concepts•first identify abstract concepts, then continue with specialization
2
Example: Travel Ontology• Transportation is conceptualized with a Top-down strategy
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
35
Methode of Uschold and Kingidentifypurpose capture coding integrating documen-
tationevaluation
Building
Ontologie Building•Ontology Capture
•Top-down identification of ontology concepts
2
object
concrete object abstract object
is-subClass-of
Subway City Bus Taxi Transportwith Taxi
Transportwith Bus
Transportwith Subway
usesuses
uses
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
36
Methode of Uschold and Kingidentifypurpose capture coding integrating documen-
tationevaluation
Building
Ontologie Building•Ontology Capture
•Top-down identification of ontology concepts
•Level of detail can be better controled•perhaps abstract concepts are not needed at all for ontology / application
•Less stability of the model → then revision is necessary (even more expensive)
2
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
37
Methode of Uschold and Kingidentifypurpose capture coding integrating documen-
tationevaluation
Building
Ontologie Building•Ontology Capture
•Middle-Out identification of ontology concepts•Start with core concepts, then specialication and/or generalization
2
Example: Travel Ontology• Transportation is conceptualized with a Middle-Out strategy
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
38
Methode of Uschold and Kingidentifypurpose capture coding integrating documen-
tationevaluation
Building
Ontologie Building•Ontology Capture
•Middle-Out identification of ontology concepts
2
Subway Bus Taxi
Transportation
is-subClass-of
City Bus Shuttle Bus Travel Bus
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
39
Methode of Uschold and Kingidentifypurpose capture coding integrating documen-
tationevaluation
Building
Ontologie Building•Ontology Capture
•Middle-Out identification of ontology concepts
•well balanced (wrt. level of detail / abstraction)•more stable than the other two methods
2
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
40
Methode of Uschold and Kingidentifypurpose capture coding integrating documen-
tationevaluation
Building
Ontologie Building•Coding
•All who participate in the development of the ontology must agree on a common structure of the knowledge base
•Integration of Existing Ontologies•Decision, whether and how existing ontologies should be reused
•can be performed in parallel with the other activities
2
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
41
Methode of Uschold and Kingidentifypurpose capture coding integrating documen-
tationevaluation
Building
Evaluation •Technical evaluation of the ontologies and the application software in each step of the development process
Documentation•Establishing guidelines for documentation
3
4
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
42
4.Ontology Engineering4.1.Ontologies revisited4.2.Methodologies of Ontology Design
4.2.1. In General
4.2.2. Method of Uschold and King4.2.3. Ontology 1014.2.4. Unified Process for ONtology (UPON)
4.2.5. Ontology Design Patterns
4.3.Ontology Learning4.4.Ontology Mapping and Ontology Merging
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
43
Ontology Development 101 (Noy, McGuinness, 2000)
•Example of a wine and food ontology
A sharedontology on
wine and food
Which wine fis the right one for fish?French wine-growing
regions and wines
Californian wine-growing regions
and wines
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
44
Ontology Development 101 (Noy, McGuinness, 2000)
•Example of a wine and food ontology
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
45
Ontology Development Process
•in practice an iterative Process that repeats continuously and improves the ontology
•there are always different approaches for modelling an ontology
•in practice the designated application decides about the modelling approach
determinescope
considerreuse
enumerateterms
defineclasses
defineproperties
defineconstraints
createinstances
„There is no one correct way tomodel a domain. There are always viable alternatives.“
Dienstag, 15. Januar 13
•Which domain should be covered by the ontology?
•What should the ontology be used for?
•What types of Questions should be answered by the knowledge represented in the ontology?
•Who will use and maintain the ontology?
•Formulation of Competence Questions
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
46
Determine Domain and Focus
determinescope
considerreuse
enumerateterms
defineclasses
defineproperties
defineconstraints
createinstances
These Questions might change within the ontology life cycle
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
47determine
scopeconsider
reuseenumerate
termsdefine
classesdefine
propertiesdefine
constraintscreate
instances
Competence Questions (Wine Ontology)•Which properties of the wine should be considered for modelling?
•Is Bordeaux a white wine or a red wine?
•Does a Sauvignon Blanc match with fish?
•Which wine matches best for grilled meat?
•Which properties of a wine do influence whether it matches with a specific dish?
•Does the bouquet of a wine change with different vintages?
•...
Determine Domain and Focus
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
48
Consider Reuse
determinescope
considerreuse
enumerateterms
defineclasses
defineproperties
defineconstraints
createinstances
•Why should we consider reuse?•in order to save cost•in order to apply tools that are applied for other existing ontologies also for our own ontology
•in order to reuse ontologies that have been validated by their application
If you don‘t find a suitable ontology or if the adaption is too complex then create a new ontology!
Dienstag, 15. Januar 13
•About which concepts are we talking?•Which properties have these concepts?•What do we want to say about these concepts?
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
49
Develop a Terminology
determinescope
considerreuse
enumerateterms
defineclasses
defineproperties
defineconstraints
createinstances
Example: Wine Ontology•wine, grape, winery, location,... •a wine‘s color, body, flavor, sugar content,... •subtypes of wine: white wine, red wine, Bordeaux wine,...•types of food: seafood, fish, meat, vegetables, cheese,...•...
Dienstag, 15. Januar 13
•Classes are concepts in the designated domain•class of wines•class of wineries•class of red wines•...
•Classes are collections of objects with similar properties•Choose a top-down / bottom-up / middle-out approach to model class hierarchies
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
50
Develop Classes and Class Hierarchies
determinescope
considerreuse
enumerateterms
defineclasses
defineproperties
defineconstraints
createinstances
Dienstag, 15. Januar 13
•Properties in a class definition describe attributes of instances•every wine has a color, residual sugar, producer, etc...
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
51
Define Properties
determinescope
considerreuse
enumerateterms
defineclasses
defineproperties
defineconstraints
createinstances
Dienstag, 15. Januar 13
•Property constraints (restrictions) describe or restrict the set of possible property values•The name of a wine is a String•The producer is an instance of Winemaker
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
52
Define Property Constraints
determinescope
considerreuse
enumerateterms
defineclasses
defineproperties
defineconstraints
createinstances
Dienstag, 15. Januar 13
•Create Instances for the classes•Every class directly becomes the type of its instances•Every superclass of a direct type is also type of its instances
•Create instances for properties, i.e. assignment of property values for the instances according to the given constraints
•„the glass of red wine that I drank last supper...“
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
53
Definition of Instances
determinescope
considerreuse
enumerateterms
defineclasses
defineproperties
defineconstraints
createinstances
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
54
4.Ontology Engineering4.1.Ontologies revisited4.2.Methodologies of Ontology Design
4.2.1. In General
4.2.2. Method of Uschold and King4.2.3. Ontology 1014.2.4. Unified Process for ONtology (UPON)
4.2.5. Ontology Design Patterns
4.3.Ontology Learning4.4.Ontology Mapping and Ontology Merging
Dienstag, 15. Januar 13
•Based on Unified Process (UP) methodology in software development and Unified Modelling Language (UML)
•Use-Case driven, i.e. more suitable for application ontologies than for domain ontologies
•Goals:•Reduction of time and cost in the development of large scale ontologies
•Quality improvement of the developed ontology via progressive validation of intermediate results
•Methodology for efficient collaboration of Knowledge Engineers and Domain Experts with clear separation of roles
•Intermediate results can already be evaluated by the user
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
55
Unified Process for Ontology Building De Nicola, Missikoff, Navigli (2005)
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
56
Unified Process for Ontology Building De Nicola, Missikoff, Navigli (2005)
•Development is divided into Cycles, which are subdivided into 4 Phases of Iterations (Inception, Elaboration, Construction, Transition). Each iteration results in a new prototype
•Each iteration consists of 5 workflowes (Requirements, Analysis, Design, Implementation, Test)
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
57
Unified Process for Ontology Building De Nicola, Missikoff, Navigli (2005)
•Workflows and Phases are almost orthogonal, i.e. involvement of single workflows in different phases of ontology development does vary
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
58
Unified Process for Ontology Building De Nicola, Missikoff, Navigli (2005)
(1)Requirements Workflow
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
59
Unified Process for Ontology Building De Nicola, Missikoff, Navigli (2005)
(2)Analysis Workflow
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
60
Unified Process for Ontology Building De Nicola, Missikoff, Navigli (2005)
(3)Design Workflow
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
61
Unified Process for Ontology Building De Nicola, Missikoff, Navigli (2005)
(4)Implementation Workflow
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
62
Unified Process for Ontology Building De Nicola, Missikoff, Navigli (2005)
(5)Test Workflow
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
63
4.Ontology Engineering4.1.Ontologies revisited4.2.Methodologies of Ontology Design
4.2.1. In General
4.2.2. Method of Uschold and King4.2.3. Ontology 1014.2.4. Unified Process for ONtology (UPON)
4.2.5. Ontology Design Patterns
4.3.Ontology Learning4.4.Ontology Mapping and Ontology Merging
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
64
Ontology Design Patterns (Gangemi, 2005)
•Adapting an Idea originally from Architecture •recurring modeling problems•providing a set of adaptable standard solutions
•Ontology Design Patterns provide•small reusable (abstract) ontology templates with explicit documentation
•searchable repository ordered by competence questions
•We distinguish:•Content Patterns
•Domain dependent, language independent •Logical Patterns
•Domain independent, related to representation language•Presentation Patterns
•Ontology from user perspective, as e.g. naming conventions•Transformation Patterns
•how to transform an ontology in another representation language
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
65
Content Pattern vs. Logical Pattern
•Logical ODPs solve design problems independently of a particular conceptualization
•Content ODPs are patterns for solving design problems for the domain classes and properties that populate an ontology; they address content problems
•Content ODPs are instantiations of Logical ODPs (or of compositions of Logical ODPs)
•Modeling problems solved by Content ODPs have two components: domain and requirements. •the same domain can have many requirements •the same requirement can be found in different domains
•A typical way of capturing requirements is by means of competency questions
•Content ODPs are collected and described in catalogues and comply to a common presentation template
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
66
Content Pattern - A Simple Example
•Example: taking over a temporary role•e.g.: Basil Rathbone played Sherlock Holmes in the 1939 movie „The Hound of the Baskervilles“
•Analyze the sentence, detect the modeling issues, and match to the Content ODPs
•A person plays a character •represent objects and
the roles they play
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
67
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
684.Ontology Engineering
4.1.Ontologies revisited4.2.Methodologies of Ontology Design
4.2.1. In General
4.2.2. Method of Uschold and King4.2.3. Ontology 1014.2.4. Unified Process for ONtology (UPON)
4.2.5. Ontology Design Patterns
4.3.Ontology Learning4.4.Ontology Mapping and Ontology Merging
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
69
Ontology Learning•Ontology Design is very expensive wrt. time and resources•can we automate the process or at least some parts?
•Ontologies can be „learned“ automatically
•Ontology Learning defines a set of methods and techniques•for fundamental development of new ontologies •for extension or adaption of already existing ontologies
•in a (partly) automated way from various resources.
•also referred to as Ontology Generation, Ontology Mining, or Ontology Extraction
•Automatisation requires help from•Natural Language Processing (NLP)•Data Mining•Machine Learning techniques (ML)
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
70
Ontology Learning•Fundamental data for ontology learning:
•Structured Data
•Semi-structured Data
•Unstructured Data
XML
HTMLXML
HTML
Machine Learning
Natural Language Processing
+Machine Learning
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
71
Ontology Learning - Basic Approach (I)
…
document corpus
(1) Term extraction <dog> <dogs>
<cat> <siamese cat>
terms ontology
(2) Conceptualizationpet
dog
siamese cat
cat
(3) Evaluation and Adaption
semi-automated process
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
72 • Natural Language Processing:(1)Tokenizer / Sentence Splitter(2)Morphological Analysis
• Stemming
• Lemmatizer(3)POS-Tagger
• Syntactic categories (verb, nomen, preposition, etc...)
(4)Regular Expression Matching(5)Chunks
• Detection of larger coherent structures in sentences(6)Syntactic Parser
Ontology Learning - Basic Approach (II)
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
73
Ontology Learning - Layer Cake
Termsriver, country, nation, city, capital, ...
Multilingual Synonyms{country, nation, Land}
Concept Descriptionc:=country:=<description(c), uri(c)>
Hierarchy of Conceptscapital ⊑c city , city ⊑c InhabitedGeoEntity
RelationsflowThrough(dom:river, range:GeoEntity)
Hierarchy of RelationscapitalOf ⊑R locatedIn
Axiomatic Schematariver ⊓ mountain = ∅
General Axions∀x((country(x)→∃y capitalOf(y,x)) ∧ (∀z (capitalOf(z,x)→y=z)))
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
74
Ontology Learning Tasks
•which tasks from ontology development can be automated?
Ontology Learning Tasks •Ontology creation•Ontology schema extraction•Extraction of ontology instances•Ontology integration and navigation•Ontology update•Ontology enrichment
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
75
Ontology Learning Tasks
•Ontology creation•Design from the scratch by an expert•Maschine Learning (ML) supports the expert during design by
•Suggestions of well suited relations among concepts•Integrity / consistency checking of the designed ontology
•Ontology schema extraction•Extraction of schemata from web documents / text documents / etc.
•ML uses input data and meta ontology to create fully-fledged domain ontologies (with the help of human experts)
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
76
Ontology Learning Tasks
•Extraction of ontology instances•Extraction of ontology instances from semi-structured or unstructured data to fill already existing ontology schemata with individuals
•applies technologies from Information Retrieval and Data Mining
•Ontology integration and navigation•Reconstruction of existing knowledge bases and navigation in existing knowledge bases,
•e.g. translation of a knowledge base from FOL to OWL
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
77
Ontology Learning Tasks
•Ontology update•Extension, reconstruction and adaption of already existing ontologies, e.g. adation to a changed domain
•relates to parts of ontologies that have been created in the way that they can be changed
•Ontology enrichment•(also Ontology tuning) relates to automated update of smaller parts of existing ontologies
•doesn‘t changes concepts and relations, but refines them (more precise)
•in difference to ontology update only parts of the ontology are considered that usually shouldn‘t be changed
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
78
4.Ontology Engineering4.1.Ontologies revisited4.2.Methodologies of Ontology Design
4.2.1. In General
4.2.2. Method of Uschold and King4.2.3. Ontology 1014.2.4. Unified Process for ONtology (UPON)
4.2.5. Ontology Design Patterns
4.3.Ontology Learning4.4.Ontology Mapping and Ontology Merging
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
79
Ontology Mapping
•to communicate, partners must use the same formal specification of a formal conceptualization
•but, to agree on the same ontologyis not always a simple task...(different applications, different views and opinions, different contexts,...)
•Partners, who use different ontologies (for the same domain) will not be able to communicate or to understand each other
•Ontologies must be mapped upon each other(= Ontology Mapping / Ontology Matching / Ontology Alignment )
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
80
Ontology Mapping
•is a process, where two ontologies are set in relation with each other on a conceptual level (Schema Matching).
•Thereby, instances of the start ontology OS will be transformed into instances of the target ontology OT according to their semantical relationship by using a mapping M: OS → OT .
•The mapping M can be•injective (not invertible) or also•bijective sein
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
81
Heterogeneity of Ontology
•Syntactical Heterogeneity:•Ontologies are available in different ontology representation languages
(E.g.: in OWL DL and F-Logic)•can be resolved on the conceptual level, most times preserving the
semantics
•Terminological Heterogeneity:•Naming differences for the identification of entities in different
ontologies (E.g.: ,Artikel‘ and ,Publication‘)•Might occur because different (natural) languages are used
•Conceptional (Semantic) Heterogeneity•Ontologies model the same domain, but in different ways•Differences might occur in completeness, granularity, perspective, etc.
•Semiotic (Pragmatic) Heterogeneity•Differences in interpretation of the domain to be modelled by humans
(difficult)
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
82
Ontology Mapping
Thing
Car Locomotive
Big Car
Bus
Engine
Cylinder Horsepower
is_a
is_a
is_a
is_a
has
hashas
has
Object
Wheeled
Train Automobile
AutobusHorsepower
is_a
is_ais_a
is_ahas has
Ontology 2Ontology 1
owl:equivalentClass
owl:equivalentClass
owl:equivalentClass
owl:equivalentClass
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
83
Ontology Mapping
•Ontology Mapping is not a “new” problem…•the same problems occur in Data Integration, e.g. for federated databases
•Federated databases manage local schemata for each contributing local database
•Data Integration (Schema Matching) can be applied via•bilateral mapping or•global Schemata and a mapping for each single local schema (mapping can be implemented via view)
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
84
Ontology Mapping Process
•Basic approach
OS1
OS2
importontologies
findsimilarities
specifymapping / merging
mappingM(OS1)
mergedontology
1 2 3
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
85
Ontology Mapping Process
Schema-Based Matching Techniques70
Structure-LevelElement-Level
Syntactic External Syntactic External Semantic
String-Based• Name
similarity
• Description similarity
Language-Based• Tokenization
• Lemmati-zation
• Morphology
LinguisticResources• Lexicons
• Thesauri
Constraint-Based• Type
similarity
• Key properties
AlignmentReuse• Entire
schema or
• Ontology fragments
Upper LevelvsDomain specificOntolo-gies
Data Analysis & Statistics• Frequency
• Distribution
Graph-Based• Graph
homo-morphism
• Path, children,
Taxonomy-Based• Taxonomy
structure
RepositoryOfStructures• Structure
metadata
Model-Based• DL reasoner
• SAT Solver
Linguistic Internal Relational
Terminological Structural Semantic
Schema-Based Matching Techniques
Extensional
Schema-Based Matching Techniques
Kind of Input
Basic Techniques
Granularity /Input Interpretation
Euzenat, Shvaiko: Ontology Matching, Springer 2007Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
86
Ontology Merging
70•is a process, where from two or mor start ontologies a new ontology is created.
•The new ontology unifies and substitutes the original start ontologies.
•Union ApproachThe new ontology is the union of all entities of the start, where conflicts arrising from different representations of identical concepts from the start ontologies must be resolved.
•Intersection Approach (extensional)The new ontology only consists out of parts of the start ontologies that overlap.
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
87
4.Ontology Engineering4.1.Ontologies revisited4.2.Methodologies of Ontology Design
4.2.1. In General
4.2.2. Method of Uschold and King4.2.3. Ontology 1014.2.4. Unified Process for ONtology (UPON)
4.3.Ontology Learning4.4.Ontology Mapping and Ontology Merging
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
Linked
Data
Applic
ations
&
Semant
ic Sea
rch
Next lecture
88
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
89
4. Semantic Web Technologies4.1 Ontological Engineering
Literature
» A. Gomez-Perez et al.Ontological Engineering, Springer, 2004.
» J. Euzenat, P. Shvaiko: Ontology Matching, Springer, 2007.
Dienstag, 15. Januar 13
Vorlesung Semantic Web, Dr. Harald Sack, Hasso-Plattner-Institut, Universität Potsdam
90
□Bloghttp://semweb2013.blogspot.com/
□Webseitehttp://www.hpi.uni-potsdam.de/studium/lehrangebot/itse/veranstaltung/semantic_web_technologien-3.html
□bibsonomy - Bookmarkshttp://www.bibsonomy.org/user/lysander07/swt1213_12
4. Semantic Web Technologies4.1 Ontological Engineering
Dienstag, 15. Januar 13