building an ontological base for experimental evaluation of semantic web applications peter...
Post on 23-Dec-2015
216 Views
Preview:
TRANSCRIPT
Building an Ontological Base for Experimental
Evaluation of Semantic Web Applications
Peter Bartalos, Michal Barla, Gyorgy Frivolt, Michal Tvarožek,
Anton Andrejko, Mária Bieliková and Pavol Návrat
{name.surname}@fiit.stuba.skInstitute of Informatics and Software Engineering
Faculty of Informatics and information Technologies Slovak University of Technology in Bratislava
Motivation
• Semantic Web applications Experimental Evaluation (SWEE)– Semantic annotation of the information
– Searching in semantic information space
• AKTORS
• Knowledge Web
• On-To-Knowledge
• NAZOU – job offers (nazou.fiit.stuba.sk) Tools for acquisition, organization and maintenance of knowledge in an
environment of heterogeneous information resources
• MAPEKUS – scientific publication (mapekus.fiit.stuba.sk)Modeling and Acquisition, Processing and Employing Knowledge About User Activities in the Internet Hyperspace
• Demand for well-built large scale ontologies with specific properties
• Filling the ontology with instances (not it’s creation) – building the A-box
Outline
• Approaches to ontological base creation
• Method for ontological test base building
• Evaluation
• Conclusions
Filling the ontology with instances
jo:JobOffer
jo:hasPrerequisite*
jo:offersPosition
jo:Organization
jo:isOfferedByjo:isOfferedVia
jo:hasBenefit*
jo:hasApplyInformation*
r:Region
jo:hasDutyLocation*
jo:Salary
jo:hasSalary
jo:offers*jo:mediates*
jo:isBasedAt
r:isPartOf* r:consistsOf*
jo:ApplyInformation
c:ProfessionClassification
jo:Prerequisite
jo:Benefit
Web
3) Wrappers
1) Generic editor
4) Generators
2) Specialized editor
SWEE OntologyFile HelpEdit
Profession
Clerks
Armed Forces
Lorem
Lorem
lor
Enter Text
Edit
lor
Lorem IpsimSaldksa;d saD sadSasadsa d’sadsaSalkdjlkfjd fld;skfDlks ljflkdsjfdsf kldsf dsdskjfldsk fj
Edit
Specialized editorSpecialized editor
jo:JobOffer
jo:hasPrerequisite*
jo:offersPosition
jo:Organization
jo:isOfferedByjo:isOfferedVia
jo:hasBenefit*
jo:hasApplyInformation*
r:Region
jo:hasDutyLocation*
jo:Salary
jo:hasSalary
jo:offers*jo:mediates*
jo:isBasedAt
r:isPartOf* r:consistsOf*
jo:ApplyInformation
c:ProfessionClassification
jo:Prerequisite
jo:Benefit
Generic editorGeneric editor
Profession
Clerks
Armed ForcesLorem Edit
Manual approaches
Automatic approaches
Generic ontology editors
• Understand the generic structure of the ontology
• Immediately usable
• Domain independent
• Insufficient validation and user comfort
• Suitable for experts (ontology engineers)
jo:JobOffer
jo:hasPrerequisite*
jo:offersPosition
jo:Organization
jo:isOfferedByjo:isOfferedVia
jo:hasBenefit*
jo:hasApplyInformation*
r:Region
jo:hasDutyLocation*
jo:Salary
jo:hasSalary
jo:offers*jo:mediates*
jo:isBasedAt
r:isPartOf* r:consistsOf*
jo:ApplyInformation
c:ProfessionClassification
jo:Prerequisite
jo:Benefit
Generic editorGeneric editor
Profession
Clerks
Armed ForcesLorem Edit
Generic ontology editors
Specialized ontology editors
• Freedom in adjusting to a given ontology and user requirements
• Sophisticated validation
based on the knowledge
of the ontology
• Development and maintenance costs
• Coupled to a ontology
• Suitable also for non-experts
File HelpEdit
Profession
Clerks
Armed Forces
Lorem
Lorem
lor
Enter Text
Edit
lor
Lorem IpsimSaldksa;d saD sadSasadsa d’sadsaSalkdjlkfjd fld;skfDlks ljflkdsjfdsf kldsf dsdskjfldsk fj
Edit
Specialized editorSpecialized editor
Specialized ontology editors
JOE – Job Offer Editor
Wrappers
• Parse Web pages and produce structured output
• Need well structured pages
• Do not need a human involvement
• Significant amount of acquired data
• Development and maintenance costs
Generators
• Reusing the already existing data
• Increase the size of the ontological base
• Instances of desired properties
• Development and maintenance costs
• Meaningfulness of the data
Approaches to Ontological Base Creation
• Different approaches have different benefits and disadvantages
• They support each other
• They can be adjusted
– Invested time
– Development of tools
Method of Ontological Base Creation
• Specification of the requirements for the ontology
– Amount of data
– Range of properties of the instances
– Instance detail
– Quality
• Analysis of the domain and information sources
• Generally no approach can separately satisfy the requirements
• Adjusting the manual and automatic approaches
Method of Ontological Base Creation
Web
) 3 Wrappers
SWEE Ontology
1jo:JobOffer
jo:hasPrerequisite*
jo:offersPosition
jo:Organization
jo:isOfferedByjo:isOfferedVia
jo:hasBenefit*
jo:hasApplyInformation*
r:Region
jo:hasDutyLocation*
jo:Salary
jo:hasSalary
jo:offers*jo:mediates*
jo:isBasedAt
r:isPartOf* r:consistsOf*
jo:ApplyInformation
c:ProfessionClassification
jo:Prerequisite
jo:Benefit
Generic editorGeneric editor
Profession
Clerks
Armed ForcesLorem Edit
) Generic editor
editor2) Specialized
File HelpEdit
Profession
Clerks
Armed Forces
Lorem
Lorem
lor
Enter Text
Edit
lor
Lorem IpsimSaldksa;d saD sadSasadsa d’sadsaSalkdjlkfjd fld;skfDlks ljflkdsjfdsf kldsf dsdskjfldsk fj
Edit
Specialized editorSpecialized editor 4) Generators
Satisfaction of the requirements to ontological data
Amount Range ofproperties
Details Resamblence toreal data
Sat
isfa
ctio
n o
f re
qu
irem
ents
Satisfaction of the requirements to ontological data
• Generic editor
Amount Range ofproperties
Details Resamblence toreal data
Sat
isfa
ctio
n o
f re
qu
irem
ents
Satisfaction of the requirements to ontological data
• Generic editor
• Specialized editor
Amount Range ofproperties
Details Resamblence toreal data
Sat
isfa
ctio
n o
f re
qu
irem
ents
Satisfaction of the requirements to ontological data
• Generic editor
• Specialized editor
• Wrappers
Amount Range ofproperties
Details Resamblence toreal data
Sat
isfa
ctio
n o
f re
qu
irem
ents
Satisfaction of the requirements to ontological data
• Generic editor
• Specialized editor
• Wrappers
• GeneratorsAmount Range of
propertiesDetails Resamblence to
real data
Sat
isfa
ctio
n o
f re
qu
irem
ents
Evaluation of the method
• NAZOU (nazou.fiit.stuba.sk)
– Ontology consists of 740 classes (670 belong to taxonomies)
– All approaches used
• MAPEKUS (mapekus.fiit.stuba.sk)
– Ontology consists of 390 classes (360 belong to taxonomies)
– Only one approach used
Conclusions
• Solution for building ontologies for semantic Web application experimental evaluation
• Tunable method based on different approaches of ontology instance creation
• Evaluated in the domain of job offers and scientific publication
• Developed two SWEE ontologies
– Job offer ontology
– Publication metadata ontology
top related