SEAL A Framework for Developing SEmantic PortALs

1
SEAL A Framework for Developing SEmantic PortALs

• Nenad Stojanovic (1), Alexander Mädche (2),
• Steffen Staab (1,3),Rudi Studer(1,2,3),
• Y. Sure (1)
• Institute AIFB, University of Karlsruhe
• http//www.aifb.uni-karlsruhe.de/WBS
• FZI Research Center on Information Technologies,
  Karlsruhe
• http//www.fzi.de/wim
• ontoprise GmbH, Karlsruhe
• http//www.ontoprise.de
• K-CAP 2001
• Victoria, B.C., Canada
• October 2001

2
Agenda

• A Framework for Developing SEmantic PortALs -
  SEAL
• Case Study - AIFB Web

3
1. Motivation Semantic Web The Vision

• Web is already an impressive success
• amount of available information
• number of human users
• However Web is currently only for human readers,
  but The Semantic Web 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)

4
Semantic Web Goals and Requirements

• Make Web information practically processible by a
  computer
• gt Making the Web more effective for its users
  by
• building an open platform for integrating data,
• providing precise searches,
• automating retrieval of information,
• drawing conclusions from information found in two
  or more separate sources, etc.
• Machine-understandable semantics of information

gt Ontologies!
5
2. Semantic Portals Portals - Windows into the
Web

• Web Portals
• structured access to large volumes of
  unstructured information
• central source of information
• brochuring the web

• Community Web Portals
• Web Portal
• homogeneous group of users (similar
  interests/goals)
• community provides, uses and exchanges
  information
• common agreement achieved in the community

6
Semantic Portals - Windows into the Semantic Web

• Semantic Community Web Portal - Semantic Portal
• Community Web Portal
• uses semantic means to structure information
• provides multiple views to information
• information are accessible to human and software
  agents on a semantic basis
• gt Ontology as a core component!

7
Semantic Portals - Development process
Feasibility study
Ontology
Requirements specification
Web site design
Implementation
Maintenance
Ontology-supported lifecycle of a Web Application
Addapted from Fraternali, P., Tools and
approaches for developing data-intensive Web
applications a survey, ACM Computing Surveys,
Vol.31, No.3, pp 227-263", 1999
8
Semantic Portals - Development process
Feasibility study
Ontology
Requirements specification
Web site design
Implementation
Maintenance
Ontology-supported lifecycle of a Web Application
Addapted from Fraternali, P., Tools and
approaches for developing data-intensive Web
applications a survey, ACM Computing Surveys,
Vol.31, No.3, pp 227-263", 1999
9
The SEAL Approach Vision

• Put the Semantic Web into practice for
  communities of interest
• reflect basic paradigm of the Webself-organizati
  on
• present a structured semantic view onto the Web
  or intranet
• crucial aspect to enable sharing of the content
  between agents
• Ontologies provide the required conceptualizations

10
SEAL - Framework for developing SEmantic portALs
Feasibility study Common- KADS
Ontology OTK Meth.
Requirements specification
Web site design
Implementation
Maintenance
SEAL-supported lifecycle of a Semantic Portal
11
SEAL Ontology development
Maintenance Evolution
Ontology Kickoff
Feasibility study
Refinement
Evaluation

• Requirement specification
• Analyze knowledge sources
• Develop baseline ontology

• Check requirements
• Test in target application
• Analyze usage patterns
• Deployment

• Knowledge elicitation with domain experts
• Develop and refine target ontology

• Manage organizational maintenance process (Who is
  responsible? How is it done?)

• Identify people
• Focus domain
• GO / No GO decision

Developed in EU-IST project On-to-Knowledge
12
SEAL Portal design - general architecture
User Interface
Presentation Engine
Providing
Accessing
Processing
Services
Inference Engine
Data
13
SEAL Web-site design

• Providing information by community users

• Ontology-based templates
• Designed according to underlying ontological
  structures (concepts, relations, axioms)
• valid information
• no redundant entering
• Easy-to-use interface for
• providing users data and
• converting them into metadata
• Example Using axioms from domain ontology for
  designing templates

14
Onto-based templates
15
SEAL Web-site design

• Portal access by human agents

• Ontology-based navigation
• Hyperlinks represent semantic relations as
  defined by the ontology

16
Onto-based navigation
Hyperlinks to pages, which corresponds to the
relations of the Person concept in the given
ontology
17
SEAL Web-site design

• Portal access by human agents

• Ontology-based navigation
• Hyperlinks represent semantic relations as
  defined by the ontology

• Ontology-based querying
• query interfaces compiled according to the
  ontology and/or the knowledge base
• deliver integrated answers extended by derived
  facts (inferencing)

18
Onto-based querying
Relations defined in the Ontology
Filled out from Knowledge base
19
SEAL Web-site design

• Portal access by human agents

• Ontology-based navigation
• Hyperlinks represent semantic relations as
  defined by the ontology

• Ontology-based querying
• query interfaces compiled according to the
  ontology and/or the knowledge base
• deliver integrated answers extended by derived
  facts (inferencing)

• Postprocessing Semantic ranking
• Delivered answers are ranked according tothe
  semantic similarity of the query result and the
  underlying knowledge warehouse

20
Semantic Ranking

• Use the characteristics of the ontology and
  associated KB to provide semantically ranked
  presentation of results
• Example

KB

• member_of(Raphael, OntoWeb), has_area(OntoWeb,
  KP)
• member_of(York, OTK), has_area(OTK, KMS)
• ...

...
Query Give me all experts working in KM area 1.
Raphael 1. York 2. York 2. Raphael
or
21
Onto-based Ranking
Semantic Ranking of retrieved answers
22
SEAL Web-site design
Portal access by machine agents

• RDF Generator
• dynamically generates RDF statements on each of
  the static and dynamic pages of the semantic
  portal
• RDF statements are generated according to the
  underlying ontology
• RDF Crawler
• used by software agents to collect RDF annotated
  information from dedicated portion of the Web
• it is a tool which downloads interconnected
  fragments of RDF from the internet and builds a
  knowledge base from this data

23
SEAL Portal design
Portal access by machine agents
gt from a Semantic Portal to the Semantic Web
24
SEAL - AIFB Case Study
Portal access by machine agents
gt from a Semantic Portal to the Semantic Web
25
3. Case Study - AIFB Web

• Semantic Portal of AIFB Institute
• http//aifb.uni-karlsruhe.de
• Semantic backbone AIFB Ontology
• Provide semantic access to
• people,
• projects and
• research area information
• Provide access to students and researches
  services
• teaching/exams
• reports
• events

26
AIFB Web - Implementation
Web Serv. Apache
Presentation Engine
JAVA Servlets
Providing
Accessing
Processing
Inference Engine
Onto- broker
RDBMS
27
AIFB Web - AIFB Ontology

• Developed in a collaborative process of AIFB
  staff
• according to presented OTK Methodology for
  Ontology Development
• using ontology environment - OntoEdit
• simple centralized ontology is currently offered
• around
• 170 concepts
• 75 relations
• 6 axioms
• sub-topic hierarchy of researches topics
• still first version
• maintenance phase will come soon

28
Onto-based templates
29
Onto-based querying
Relations defined in the Ontology
Filled out from Knowledge base
30
Onto-based navigation
Hyperlinks to pages, which corresponds to the
relations of the Person concept in the given
ontology
31
Onto-based Ranking
Semantic Ranking of retrieved answers
32
Conclusion

• SEAL is a comprehensive methodology for realizing
  Semantic Web vision
• Ontologies provide the semantic underpinning for
  framework for developing SEmantic portALs
• Maintenance of ontologies is very important phase
  in semantic portal life cycle
• exploit available resources gt ontology learning
  and information extraction

33
Thank you!

• www.aifb.uni-karlsruhe.de
• www.fzi.de/wim
• www.ontoprise.de

34
Semantic Portals - Development process
Lifecycle of a Web Application
Source Fraternali, P., Tools and approaches for
developing data-intensive Web applications a
survey, ACM Computing Surveys, Vol.31, No.3, pp
227-263", 1999
35
CommonKADS - Context Modeling Road Map
Start
OM-1 worksheet problems, solutions, context
TM-1 worksheet task analysis
OM-3 worksheet process break-down
If feasible
TM-2 worksheet knowledge item analysis
Focus domain for ontology development
OM-5 work-sheet Judge Feasibility (Decision
Document)
OM-2 worksheet description of organi-zation
focus area
Refine
Integrate
OM-4 worksheet knowledge assets
AM-1 worksheet agent model
If unfeasible
Stop
36
Ontology Requirements Specification Document
Ontology Requirements Specification Document
Name Research-interest Ontology
Date 2001/06/06 Ontology Engineer T. Model

• Domain and Goal
• The ontology is modelled for the domain
  research-interest which is a part of the research
  organisation
• The ontology serves as a model for sharing
  knowledge about research interests in a
  organisation
• Ontology serves as a base for semantic search for
  researcher/projects according to their research
  interests
• Design Guidelines
• The ontology contains lexical entries in the
  domain of research-interest in a research
  organisation. Research-interest are connected to
  current-preferences, educational background and
  working-projects of a person. Research-topic
  hierarchy should be modelled in more details and
  should not exceed 100 topics. Topic-hierarchy
  should be modelled at instance level (topics are
  instances and not concepts). Axioms are planned.
• Supported Applications
• Intranet based Research-skill Management System
  at institute AIFB
• Knowledge Sources
• AIFB web site - personal pages
• Research reports, publications
• Internal document about institute (organisation,
  staff)
• Interviews with researcher
• Users and Use cases
• G. Peoplefind, Human Resource Department
  attached use case 1

37
Competency Questionnaire
Competency Questionnaire No. 1
Name skill-man-ontology Date 2001/03/22 Ontology
Engineer T. Model Domain Expert X. Pert
Type
Lexical Entries
Competency Question
No.
researcher
concept
Q1
What are the name, position, telephone and e-mail
of researcher from research group EA that has
research interest in research topic Knowledge
Management and work on project with the name
Ontobroker ?
concept
research group
concept
research topic
Knowledge Management is a research topic
isA relation
researcher has research interest in research
topic
relation
isA relation
EA is a research group
relation
Researcher has name
relation
Researcher has position
relation
Researcher has telephone
...
relation
Researcher has e-mail
38
Competency Questionnaire
Competency Questionnaire No. 1
Name skill-man-ontology Date 2001/03/22 Ontology
Engineer T. Model Domain Expert X. Pert
Type
Lexical Entries
Competency Question
No.
researcher
concept
Q1
What are name, position, research-group, room
number, telefon, email for selected person
concept
research group
concept
research topic
Knowledge Management is a research topic
isA relation
researcher has research interest in research
topic
relation
isA relation
EA is a research group
relation
Researcher has name
relation
Researcher has position
relation
Researcher has telefon
...
relation
Researcher has e-mail
39
Competency Questionnaire
Competency Questionnaire No. 1
Name skill-man-ontology Date 2001/03/22 Ontology
Engineer T. Model Domain Expert X. Pert
Type
Lexical Entries
Competency Question
No.
professor
concept
Q1
What is secretary for selected professor
isArelation
Professor isA researcher
concept
secretary
relation
Professor has secretary
HiWi
Who is supervisor for a selected HiWi
concept
isArelation
HiWi isA student
relation
HiWi has supervisor
...
40
Competency Questionnaire
Competency Questionnaire No. 1
Name skill-man-ontology Date 2001/03/22 Ontology
Engineer T. Model Domain Expert X. Pert
Type
Lexical Entries
Competency Question
No.
project
concept
Q1
What are the code and name of project from
research group EA that is in topic Knowledge
Management and that has a researcher named
Carter?
concept
research group
concept
research topic
Knowledge Management is a research topic
isA relation
relation
project has research topic
isA relation
EA is a research group
relation
Project has name
relation
Project has code
relation
Researcher has name
...
41
Competency Questionnaire
Competency Questionnaire No. 1
Name skill-man-ontology Date 2001/03/22 Ontology
Engineer T. Model Domain Expert X. Pert
Type
Lexical Entries
Competency Question
No.
project
concept
Q1
What are code, name, research-group,
starting-date, description, founded by property,
researcher and research-topic for selected
project
relation
project has code
relation
project has name
relation
project has research group
relation
project has starting-date
relation
project has description
relation
project has founding
concept
project has researcher
concept
project has research-topic
...
42
Initial lexicon
Researcher Resersch-group Research-topic ... Stude
nt, Professor, ... has research group has
research topic has researcher has project .
43
Refinement Phase
STEP 1 Baseline ontology
Concepts
Researcher Resersch-group Research-topic, ....
Hierarchy
Person
Researcher
Student
Professor
SciReseacher
44
Refinement Phase
STEP 2 Seed ontology
Relations Researcher has research interest
Research-topic Project has research topic
Research-topic Project has researcher
Researcher Researcher has project
Project
Attributes-characteristics of concepts
Researcher - has name - String Project - has
code - String ...
45
Refinement Phase
STEP 2 Seed ontology
Axioms Possibility to combine
relations inverse (Person - Project) transitive
(Reseach-topic) quazi-transitive possibile
syntax patterns range-equal possibile
patterns domain-equal possibile patterns
Researcher has project Project, Project has
research topic Research-topic -gt Researcher
has research interest Research-topic
Researcher has research interest
Research-topic, Project has research topic
Research-topic -gt Researcher has project
Project
Project has researcher Researcher, Project
has research topic Research-topic -gt
Researcher has research interest
Research-topic
IF ProjectX has researcher ResearcherX AND
ProjectX has research topic Research-topicZ
THEN ResearcherX has research interest
Research-topicZ
46
Refinement Phase
STEP 3 Formalize ontology

• Concepts, Relations, Axioms
• Example In F-Logic
• Inverese
• Transitive
• Composition (combination)

FORALL ProjectX, ResearcherX ResearcherXResearc
herhas project - gtgtProjectX lt-
ProjectXProject has researcher -gtgt ResearcherX.
IF ProjectX has researcher ResearcherX AND
ProjectX has research topic Research-topicZ
THEN ResearcherX has research interest
Research-topicZ FORALL ProjectX, ResearcherX,
Research-topicZ ResearcherXResearcherhas
research interest - gtgtResearch-topicZ lt-
ProjectXProject has researcher -gtgt ResearcherX
and ProjectXhas research topic-gtgt
Research-topicZ
47
Refinement Phase
STEP 3 Ontology formalised in RDF (a part)
..... ltrdfsClass rdfID"Person"gt
ltdamlrestrictedBygt
ltdamlRestrictiongt
ltdamlonProperty rdfresource"titel"/gt
ltdamltoClass rdfresource"http//www.w3.o
rg/TR/xmlschema-2/string"/gt
lt/damlRestrictiongt lt/damlrestrictedBygt
ltdamlrestrictedBygt
ltdamlRestrictiongt
ltdamlonProperty rdfresource"nachname"/gt
ltdamltoClass rdfresource"http//www.w
3.org/TR/xmlschema-2/string"/gt
lt/damlRestrictiongt lt/damlrestrictedBygt
ltdamlrestrictedBygt
ltdamlRestrictiongt
ltdamlonProperty rdfresource"vorname"/gt
ltdamltoClass rdfresource"http//www.w3
.org/TR/xmlschema-2/string"/gt
lt/damlRestrictiongt lt/damlrestrictedBygt .
.. lt/rdfsgt
48
SEAL Ontology development
ORSD Initial lexicon
Target ontology
Roll out
Ontology kickoff
Refinement
Evaluation
Maintenance

• ONTOLOGY

• Develop baseline ontology
• Knowledge elicitation with domain experts (seed
  ontology)
• Formalize (target ontology)

• Check requirements
• Test in target application
• Analyze usage patterns

• Manage organizational maintenance process

• Requirement specification
• Analyze input sources (initial lexicon)

Developed in EU-IST project On-to-Knowledge
49
AIFB Web Implementation General architecture
50
Explanation capabilities
51
Explanation capabilities
52
Explanation capabilities
Answer on Why-question could be in textual form
53
Explanation capabilities
Answer on Why-question could be in textual form -
details
One of possibile explanations
That fact should be explained
Ground facts
New fact
First level in explanation combining of ground
facts only
Second level combining of ground facts and new
facts
Third level combining of ground facts and new
facts - It is also top level solution
54
Explanation capabilities
Answer on Why-question could be in textual form
expert view
55
Explanation capabilities
Answer on Why-question could be in graphical form
- Backward starting from result
rvoforscht_in_gebiet-gtgt c_MultimediaSysteme
56
SEAL Web-site design

• Portal access by human agents

• Ontology-based navigation
• Hyperlinks represent semantic relations as
  defined by the ontology

• Ontology-based querying
• query interfaces compiled according to the
  ontology and/or the knowledge base
• deliver integrated answers extended by derived
  facts (inferencing)

• Postprocessing Semantic ranking
• Delivered answers are ranked according tothe
  semantic similarity of the query result and the
  underlying knowledge warehouse