Engineering Semantic Web Information Systems

1
Engineering Semantic Web Information Systems

• Richard Vdovjak
• Flavius Frasincar
• Geert-Jan Houben
• Peter Barna

Databases Hypermedia Group Department of
Computer Science
/dept. of mathematics and computer science
wwwis.win.tue.nl/hera
2
Overview

• Motivating example Virtual art gallery
• Hera framework, models and technologies
• Hera Back-end Integration engine
• Hera front-end Presentation generation engine
• Summary and future work

3
Virtual art gallery WIS

• Create on-the-fly exhibitions
• painters, paintings, techniques...
• Many to many content delivery
• Triggered by the user query
• Exhibits (Images) are from (online) Art
  catalogues
• Descriptions are gathered from an (online) Art
  Encyclopedia

4
We need a framework that offers

• Semantics
• Express concepts and their hierarchies
• Relationships among the concepts
• Query language that can exploit the above
• Access to several sources
• Flexible source management
• Flexible query mediation
• User/platform adaptation
• (adaptability and adaptivity)
• Automated presentation design
• Based on the user query,
• Device profiles and the browsing history

5
The Hera Framework Design Steps, Models, and
Processing Engines
6
Overview of the System
7
Conceptual Model (CM)

• Provides a uniform semantic view over different
  data sources that are integrated within a given
  Web application
• Consists of hierarchies of concepts relevant
  within the given domain, their properties, and
  relations
• Encoded in RDF(S)

8
RDF(S), RQL

• RDF(S)
• W3C standard for describing metadata
• Directed labeled graph formalism
• Formal semantics defined
• RQL

select X from XTechniquetnameXtname
where Xtname "Chiaroscuro"
9
Conceptual Model Example
10
Source Clusters

• Sources are
• Autonomous
• (Virtually) grouped to clusters based on the
  content they provide
• RDF(S), RQL capable

11
Integration Model

• IM decouples the CM and Sources
• Articulations
• actual links between the CM and the source
  ontologies
• concept/instance uniqueness
• (a part of it) serves as a query on the source
  side
• Decorations
• offer a way to rank sources within the same
  cluster
• capture explicitly designers knowledge about
  sources
• the order in which the sources are consulted is
  flexible
• open possibilities for queries with constraints
• e.g. Im interested in the answer within 1 s,
  otherwise forget it

12
Integration Model Ontology

• Path expression
• Primary node (including its ID)
• Sequence of nodes and edges
• Articulation
• Target and Source path expressions
• Decoration
• Value based ordering criterion
• e.g. ResponseTime, Reliability
• Processing instruction
• Transformers (e.g. Literal2String)

Application independent
13
Integration Model InstanceArticulation Example
Target Path Expression
cmaname
idByValue
ends
follow
starts
cmPainting
cmpicture
pe_to3
Image
target
a2_1
obtainedFrom
obtainedFrom
follow
endsL
source
starts
acPainting
acvisualized
URL
pe_from3
applies
transformedBy
srcAddress
idByValue
URL2ImageTransf
http//wwwac
actitle
Source Path Expression (query for the source)
d2_1
0.9
Decoration (Reliability)
14
Source Management
Front-end
Back-end
Conceptual Model
Access Point
User Query
Integration Model
html/smil
IM Instance
Presentation
IM Specialization
Sources
15
Query answering

• Query mediation
• For every variable in the query find
  articulations in the IM instance pool
• If there more articulations for one variable sort
  them based on the chosen decoration(s)
• Execute the source path expression queries at
  the the sources
• Perform the required processing instruction/data
  transformations
• Assemble the results

16
Application Model (AM)

• AM serves as a presentation blue-print
• Describes hypermedia aspects of the presentation.
• Captures the navigational view over the CM
• Consists of (nested) slices and slice
  relationships
• Slices - meaningful presentation units
• Associated to concepts from the CM
• Contain properties and possibly other slices
  (nesting)
• Slice relationships
• Aggregation relationships index, tour, indexed
  guided tour
• Reference relationships link with an anchor
  specified.
• Encoded in RDF(S).

17
Application Model Example
18
Adaptation/User Model

• Captures two kinds of adaptation
• Adaptability takes into account the situation in
  which the user will use the presentation (e.g.
  the browsing platform).
• Adaptivity means that the presentation changes
  itself according to the state of the users
  mind while being browsed.
• Consists of
• Device/User Profile captures static visual and
  platform preferences encoded in CC/PP.
• User Session represents the dynamic users state,
  e.g. did the user visit (learn) this slice
  (concept).
• Application and Update Rules describe the
  behavior of the presentation (e.g. conditional
  slices in AM) and keep the User Session
  up-to-date (AHAM rules).

19
Adaptation Model Example
20
Adaptation Model Syntax

• Adaptability Condition
• Adaptivity Condition

ltrdfsClass rdfIDSlice.painting.picture
sliceconditionprfImageCapableYe
sgt ltrdfsubClassOf rdfresourceSlice/gt lt/
rdfsClassgt
ltrdfsClass rdfIDSlice.painter.main
sliceconditionumTechnique lt 10gt
ltrdfsubClassOf rdfresourceSlice/gt lt/rdfsCl
assgt
21
Profile Example

• Device/User Profile (CC/PP encoding)
• Screen size 100x80, preferred language English

ltccppcomponentgt ltupUserPreferencesgt
ltupLanguagegtEnglishlt/upLanguagegt
lt/upUserPreferencesgt
lt/ccppcomponentgt lt/rdfDescriptiongt
ltrdfDescription rdfaboutProfilegt
ltccppcomponentgt ltprfHardwarePlatformgt
ltprfImageCapablegtNolt/prfImageCapablegt
ltprfScreenSizegt100x80lt/prfScreenSize
gt lt/prfHardwarePlatformgt
lt/ccppcomponentgt
22
Rendering

• XSLT code generation
• Different code generators
• HTML for PC Web browsers
• SMIL code for multimedia presentations
• WML code for WAP phone browsers

WML
XSL
HTML
XSL
ltxsltemplate matchslice-instancegt
ltTABLEgt ltxslapply-templates
select/gt lt/TABLEgt lt/xsltemplategt
ltxsltemplate matchslice-instancegt ltCARD
id_at_idgt ltxslapply-templates
select/gt lt/CARDgt lt/xsltemplategt
23
Resulting Hypermedia Presentations
HTML
WML
SMIL
24
EROS exploring the CM

• Explorer for RDFS-based OntologieS
• multiple views over the same model
• support for RQL queries construction

25
Summary
Hera

• Semantics
• Express concepts and their hierarchies
• Relationships among the concepts
• Query language that can exploit the above
• Access to several sources
• Flexible source management
• Flexible query mediation
• User/platform adaptation
• (adaptability and adaptivity)
• Automated presentation design
• Based on the user query,
• Device profiles and the browsing history

CM in RDF, RDFS
RQL
IM Articulations
IM Decorations
UM, U/P Profile (CC/PP)
AM in RDF, RDFS
26
Present Future Work

• Applying our approach in different domains
• Virtual museum
• Photo Portal
• Medical science (drug/disease ontologies)
• Optimization issues
• Authoring tools for the underlying models
• Initial correlation for Schema Integration