IBROW Industrial Board Meeting

1
IBROW Industrial Board Meeting

• Amsterdam, July 3, 2000

2
Agenda
1100-1145 welcome and introduction to
IBROW 1145-1230 the semantic web 1230-1330
lunch 1330-1415 spin-off applications
(scenario / use case) 1415-1430
break 1430-1500 demonstrations 1500-1600
feedback session (10 minutes each) 1600-1700
discussion and wrap up 1930
dinner
3
History and context

• Phase 1 of IBROW under LTR ESPRIT
• 100 KEuro, 4 partners, 1 year (1998)
• Feasibility study, proof of concept
• Phase 2 (now) under IST, Future and Emerging
  Technologies
• 1.2 MEuro, 7 partners, 3 year (2000-2003)

4
Partners

• SWI, UvA
• KMI, OU
• IIIA, CSIC
• VU (Free University Amsterdam)

• SMI, Stanford
• Intelligent Software Components (iSOCO)
• AIFB, UKa
• Industrial advisory board

5
Industrial advisory board

• AIAI (UK)
• Bolesian (NL)
• DaimlerChrysler (DE)
• Deutsche Telekom (DE)
• IBM (JP)
• Unilever (NL)

6
Your interest in IBROW
DaimlerChrysler "The results developed by IBROW
are likely to be useful for research projects at
DaimlerChrysler dealing with knowledge-based
engineering and systematic reuse of software
engineering experiences."
Deutsche Telekom "One of our special interests
in the IBROW project is further the methodology
for representing components because it seems to
be one crucial factor for configuring new
systems. While usually the configuration of
hardware components is easy to realise, in the
sense that input/output behaviour and resource
descriptions are given by product sheets, soft
components like software, teaching material,
documents and annotations, and problem-solving
methods are by far harder to describe."
Bolesian "As a software development house
specialised in knowledge-based systems, Bolesian
considers the technology developed in the IBROW
project important for future applications of
knowledge-system technology."
7
Your interest in IBROW, cont.
Unilever "The Knowledge Mapping and Structuring
Unit at Unilever Research is particularly
interested in knowledge structuring, knowledge
reuse and libraries of reusable knowledge
components on the Web. Therefore, it is our
belief that we could benefit greatly from the
results and methods of the IBROW project."
AIAI "AIAI's interest in knowledge management,
and particularly in the representation and
distribution of organisational knowledge over the
Web, could benefit greatly from the results and
methods of the IBROW project."
IBM "The subject of the IBROW project,
intelligent brokering over WWW, is an important
research topic. Since Dr. Hori, an advisory
researcher at IBM Tokyo Research Laboratory, has
been working on technologies for reusable
software and knowledge components, I believe he
can make substantial contributions to the
advisory board."
8
Your role in IBROW
An industrial advisory board consisting of five
members from three European countries and one
from Japan oversees the scientific directions in
an economic context.
The committee will ensure the overall coherence,
technical quality and exploitability of the IBROW
project, especially in relation to industrial
needs. The steering committee will meet once a
year with the Project Manager and key project
personnel to ensure that the project is kept on
course.
9
Goal of the meeting

• Project is 5 months under way
• No yet many results
• Discussion of commercially interesting results
• Ideas for dissemination and use plan
• But, it remains a basic research project

10
IBROW

• An Intelligent Brokering Service for
• Knowledge-Component Reuse on the World-Wide Web

11
Objectives of IBROW

• Configure reasoning services on the web for users
• Third party component reuse through the Web
• Plug Play of problem solvers and knowledge
  bases

12
Example
13
Meanwhile, behind the scene
Task
Classification generate prune
Agent capabilities
find
integrate
adapt
Interoperability infrastructure
KB
KB
Program
Program
Legacy software
Web page
14
The agents involved
15
Innovative aspects

• Most web brokers handle static information
• Opens possibility for a new electronic market
  place
• Throw-away / Ad-hoc applications

16
Relation to web services
Solutions
Automatic Information Processing Support in Task
Achievement
Automatic Information Extraction Support in
Information Access
Automatic Information Retrieval Support in
Information Source Finding
17
Key issues

• Mark up of components
• Machine processable data
• Reasoning with these annotations
• Localization/matching/selection
• Negotiation
• Adaptation and integration
• Execution

18
Marking up UPML

• Capabilities of problem solvers, KB
• Goal of tasks
• Assumptions of problem solvers, tasks and domains
• Pragmatics (non functional)
• ease of use
• rate of success
• successfully used in

19
Content of markup

• Goal of classification
• Goal of prune

forall cclasses single_solution(c) lt-gt
cardinality(c)1
forall cclass, fsfeatures complete(c,f) lt-gt
(forall ffeature in(f,fs) /\ (true(c) -gt true(f))
forall xclass in(x,output) -gt in(x, input) /\
(forall pproperty in(p,properties) -gt
has_property(x,p)))
20
Syntax of markup RDF
ltrdfDescription ID"instance_00011"gt
ltrdftype resource"http//www.upml.org/upml.rdf
Formula"/gt ltformulagtsingle_solution(solution) /\
(forall sclass in(s,solution) /\
complete(s,observations)) lt/formulagt
lt/rdfDescriptiongt
ltrdfDescription ID"instance_00011"gt
ltrdftype resource"http// www.upml.org/upml.rdf
Formula"/gt ltformulagtsingle_solution(solution)
/\ (forall sclass in(s,solution) /\
complete(s,observations))lt/formulagt
lt/rdfDescriptiongt
21
Key issues

• Mark up of components
• Machine processable data
• Reasoning with these annotations
• Localization/matching/selection
• Negotiation
• Adaptation and integration
• Execution

22
Reasoning required

• Recognize the task of user
• Analyze task (requires much expertise)
• Find relevant PSs for (sub)task goals and KB
• Check applicability of PSs in KB
• Adapt and Integrate (3D-framework)
• PS--PS and KB--PS
• Ontology mapping by adaptors

23
Ontologies are key

• For common understanding

sets
class Attribute input-set
prune
class attribute value has_feature ..
24
Relevant problem solvers?

• Component matching
• selection
• keyword-based extended with thesauri
• case-based approach
• theorem-prover
• negotiation between agents
• ACL

25
Key issues

• Mark up of components
• Machine processable data
• Reasoning with these annotations
• Localization/matching/selection
• Negotiation
• Adaptation and integration
• Execution

26
Interoperability

• Protocols for combining heterogeneous components
• CORBA
• XML
• Also here ontologies play a role
• At a different level

27
Architecture 1
28
Architecture 2
Broker
Repositories
Market place
29
Open issues

• Annotation
• Difficult
• Boring
• Bridges and refiners (theory)
• Architectures
• Scale

30
IBROW versus other projects

• Metacrawler
• OntoSeek
• Shopping agents
• Jango
• Ariadne (ISI)
• OntoKnowledge
• Ontobroker

• JavaBeans By Design
• Algovista
• Retsina
• DAML
• (IBROW)

31
Increasingly intelligent solutions
Solutions
IBROW, RETSINA, DAML
Automatic Information Processing Support in
Task Achievement
Ontobroker, On-To-Knowledge, ARIADNE,
Automatic Information Extraction Support in
Information Access
Metacrawlwer, ONTOSEEK, Jango,
Automatic Information Retrieval Support in
Information Source Finding
32
Added value IBROW

• IBROW formally describes sources
• Reason automatically about them
• Enables configuration of intelligent integration
  of information by meta-level reasoning

33
Industrial seminar of phase 1

• What kind of users addressed?
• Power, non-power
• What parts are marketable now?
• Workbench functionality
• Focusing on the right knowledge?
• Domain versus problem solving
• Confidentiality of knowledge