Semantic Web for Generalized

steelsquareInternet and Web Development

Oct 20, 2013 (3 years and 11 months ago)

86 views

AIFB

1



Semantic Web for Generalized

Knowledge Management


Rudi

Studer
1, 2, 3


Siggi Handschuh
1
, Alexander Maedche
2
,

Steffen Staab
1, 3
, York Sure
1


1
Institute AIFB, University of Karlsruhe

http://www.aifb.uni
-
karlsruhe.de/WBS

2
FZI Research Center on Information Technologies, Karlsruhe

http://www.fzi.de/wim

3
ontoprise GmbH, Karlsruhe

http://www.ontoprise.de



NSF
-
EU Workshop Semantic Web

Sophia Antibolis

October 3
-
5, 2001



AIFB

2

Agenda


1.
Knowledge Process:

-
Use: KM Applications (e.g. Portals)

-
Capture: Creation and Annotation of Metadata


2.
Knowledge Meta Process

-
Ontology Learning


3.
Conclusion


Use

Capture

AIFB

3

Knowledge Meta Process
&

Knowledge Process


Knowledge Process


Working with KM Application

Knowledge Meta Process

Design, Implementation,

Maintenance

AIFB

4

Retrieval /


Access

Query

Search

Derive


Knowledge Process

Capture

Extract

Annotate


Create

Import

Documents

Metadata

Databases

Use

Apply

Summarize

Analyse

Automatic Use



Capture

Use

AIFB

5


Reduce overhead of applying KM


Seamless integration

of KM application into

working environment


Exploit

existing
legacy data
, e.g. databases



Avoid information overload


Context
-
dependent

access and presentation

of knowledge


Reflect task at hand


Reflect used output device


Personalized

access and presentation


Exploit user profile


Be able to “
forget





KM Applications

Use

AIFB

6


Anywhere and anytime access to knowledge


Intranet environment


Internet environment


Laptop/PDA/Mobile phone


Wearable devices



What you get presented


is what you need


is tailored to your profile


is adapted to the output

device



KM Applications:

Anywhere and Anytime

Use

AIFB

7



Knowledge Portals

are portals that ..



focus on the
generation
,
acquisition
,
distribution

and the
management

of
knowledge



in order to offer their users


high
-
quality access

to and


interaction possibilities

with


the contents of the portal



cf.
OntoWeb portal



Knowledge Portals

Use

AIFB

8

KAON Portal Architecture

Use

Knowledge Warehouse

Clustering

Presentation Engine

(RDF
-
)Crawler

Extractor

Browser

WWW / Intranet

Annotation

Navigation

Semantic

Query

Person
-

alization

Inference

Engine

Semantic

Ranking

AIFB

9

Use

AIFB

10

Use

AIFB

11


Exploit ontologies and related metadata


Various conceptual models are needed, a.o.


Application domain


Task at hand


User profile



Several approaches under development


Stanford’s OntoWebber


Karlsruhe’s
KAON
-
Portal


FZIBroker

as one instantiation


Integrate browsing, querying, content providing


Generating Knowledge Portals

Use

AIFB

12

Automatically Generated Portals

Use

AIFB

13

Creation and Generation

of Metadata



Manual creation of metadata for web documents is a
time
-
consuming process



Possible solutions:


Process web documents and
propose annotations

to the
annotator


Use information extraction capabilities based on simple
linguistic

methods


Exploit
domain specific lexicon

and
ontology

to bridge the
gap between linguistic and conceptual structures


Authoring

of new documents (get annotation for free)


Reuse

existing structured data, e.g. available in databases


KAON Reverse

tool

Capture

AIFB

14


Methods are currently under development in the

DAML OntoAgents

project


Cooperation project


Stanford University, DB Group (Stefan Decker)


Univ. of Karlsruhe, Institute AIFB


KAON Annotation Environment

combines


Manual creation of metadata


Semi
-
automatic generation of metadata


metadata
-
based authoring


Partially realized in the
KAON O
NT
-
O
-
MAT

tool,

available for download at


http://ontobroker.semanticweb.org/annotation/ontomat/

Creation and Generation

of Metadata

Capture

AIFB

15

KAON Annotation Environment

web
pages

domain

ontolog
ies


copy

WWW

Document
Management


Annotation

Inference

Server

Information

extraction

Component

annotate

crawl


Annotation

Tool
GUI


plugin

plugin

plugin

Ontology

Guidance

Document

Editor

Annotation Environment

query

extract

crawl



annotated

web pages

Capture

Functions
:

Knowledge Capturing

+ Annotation

Authoring

+ Annotation


AIFB

16

KAON O
NT
-
O
-
M
AT



Capturing and Annotation


Instance, relationship and attribute creation


Document markup



Authoring and Annotation


Document editing and markup


Annotation on the fly

Capture

AIFB

17

Further Issues


Semi
-
automatic generation of metadata for


Text documents


Images


Videos


Audio


Combine

multimedia standards with Semantic Web
technologies


MPEG
-
7, SMIL


RDF schema, OIL, DAML
-
OIL


Achieve
semantic interoperability

between

different standards



Capture

AIFB

18

Knowledge Meta Process for
Ontologies
(cf. OTK
-
Project)

Revision and
expansion
based on
feedback

Analyze usage
patterns

Analyze
competency
questions

ONTOLOGY

Requirement
specification


Analyze input
sources

Develop
baseline
ontology

Concept
elicitation with
domain
experts

Develop and
refine target
ontology

Manage
organizational
maintenance
process



GO


/ No GO
decision

Kickoff

Refine
-

ment

Evaluation

Main
-
tenance &

Evolution

Feasi
-


bility

Study

Ontology

Learning

AIFB

19

Ontology Learning


Lots of ontologies have to be built


Ontology engineering is difficult and time
-
consuming


Cf. tools
OntoEdit
,
Protégé
-
2000
,
OilEd



Solution:


Apply
Machine Learning

to ontology engineering


Multi
-
strategy learning


Exploit multiple data sources


Build on
shallow linguistic

analysis


Build the ontology in an application
-
oriented way, based on
existing resources


Reverse Engineering


Combine manual construction and learning into a
cooperative

engineering environment



AIFB

20

Ontology Learning:

Relation Mining

root

company

TK
-
company

Online service

company

Linguistically associated

Generate suggestion:

relation(company, company)


=> cooperateWith(company, company)

T
-
Online

Nifty

AIFB

21

Ontology Learning: Emergent
Semantics


Derive consensual conceptualizations in

a
bottom
-
up

manner



Exploit interaction in a decentralized environment


Peer
-
to
-
peer scenario


Hundreds of local ontologies


Learn alignment of ontologies
through usage


One approach within a multi
-
strategy environment



AIFB

22

Evolution of Ontology
-
based
KM Applications


Real world environment is changing all the time:


new businesses


new organizational structures in enterprises


new products and services


...


Ontologies have to reflect these changes


new concepts, relations and axioms


new meanings of concepts


concepts and relationships become obsolete



Support for evolution

of
ontologies

and
metadata


is essential


ontology
-
based applications depend on

up
-
to
-
date ontologies and metadata



AIFB

23

Conclusion



Semantic Web provides promising way for providing
relevant

knowledge


Appropriate granularity


Personalized presentation


Task
-

and location
-
aware


Reduce overhead

of …


building up and


maintaining KM applications

=>
most critical success

factor for real
-
life applications


(IT aspect)


Reduce centralization

caused by ontology
-
based
approaches


Use multiple ontologies


Combine top
-
down and bottom
-
up approaches

for ontology construction and learning



AIFB

24

KM Applications and eLearning



KM application has to be embedded into

a
learning organization



eLearning fits smoothly into such an environment


Task driven learning


Learning based on competence analysis



AIFB

25

KM Applications and eLearning



Edutella project

exploits Semantic Web framework as
a distributed query and search service

http://sourceforge.net/projects/edutella/


Peer
-
to
-
peer service for the exchange of educational
metadata


Part of
PADLR

project (Personalized Access to Distributed
Learning Repositories)


Cooperation between Stanford University and

Learning Lab Lower Saxony (L3S), Hannover, Germany

http://www.learninglab.de


Institute AIFB is Learning Lab member