Transcript Research Focus July 2002

IIT — e-Business
(Fredericton)
Bruce Spencer
Research Overview
July 10, 2002
Business Rules & Web Services
• Web Services are software components existing on server
sites across the Internet
• Available for use by clients/browsers
– remote procedure call
• Searchable registries of Web Services provide
– yellow page information (who provides what service)
– green page information (how the service can be used)
• Examples:
– stock quotes, price quotes, weather reports
– More complex: tax calculations, applications for
immigration visa
• Buy-in from IBM, Sun, Microsoft, HP, BEA, etc.
Web services and policy
• Policy dictates who receives what service
– Example: price quote depends on shoppers’
demographics and shopping history
– Applicant for immigration subject to eligibility rules
• “IF THEN” Rules for expressing policy
– If shopper has spend $500 in past year here and
item for quote is in luxury category
then discount is 10%
• Rule ensure policy is clear, maintainable, explainable
• Web services that make judgments should be based on
clear rules
– Reasons for judgments can be explained
1. Explanation of Web
Services
• When a web service judgment is questioned, client can
request
– A justification of the decision
(proof)
– A subdialog to gather more information with a chance
to offer other alternatives
(explanation and clarification)
• Example
– Client: Why not a 10% discount?
– Web Service: Have you spent more than $500 here
recently? Here is your list which totals only $200.
– Client: Add my wife’s purchases to my list.
– Web Service: OK. Now I can give you a 10% discount.
Prototype for Explanations
• Built in Java
• Server uses j-DREW (Deduction Reasoning Engine for the
Web) for making inferences
– RuleML (form of XML) for expressing rules
• Client uses Graphical User Inferface for displaying and
interacting with proofs
– User can
• Expand sub-proofs
• Question any conclusion
• Open subdialog
• Dynamically add new information
• Master’s thesis project for Keping Jia
• Planned paper in distributed declarative debugging
literature
2. Rules for Privacy Policy
• Web sites collect information about users
• APPEL from W3C is a rule-based language for
expressing a user’s privacy policy
– What personal information can web sites
take? Who can they share it with? What can
they use it for?
• Discussed on “The National” June 20, 2002,
interview with Tim Berners-Lee
Detecting APPEL rule
conflicts
• Master’s thesis project for Fang Wang
– Privacy rules can contradict each other
– A user’s private information can be protected
by a rule but, unknown to the user, that rule
can be overridden by another which allows the
information out.
– Need a tool to detect contradictory rules
– Uses 6-connective logic, non-Horn reasoning
– Publications planned for logic conference and
security conference
3. E-Procurement Planning
• Purchases of bundles of items
– each item has many options
• From whom to buy
• How much (could be probabilistically
predicted)
• Limited time offers
– Combinatorially many ways to procure bundles
• Users have different criteria, including price range,
quality, reputation, etc.
– Use multiattribute utility
• User may be risk averse or risk seeking
– Utility of spending money
Problem
• Setting: large volume of items, combinations,
predicted prices, quotes available at different
times, rescinded offers
• Need purchase a bundle
– To maximize utility both of items and of spent
money,
– To exploit the benefits of deferring choices
until many quotes are open
(expected utility of future choice set)
• PhD Thesis for Scott Buffett
• Planned two journal publications, including IJEC
3. j-DREW
• Highly configurable reasoning engine for clausal first
order logic
• Written in Java
• Programmers with basic data structure skills are able to
program new search strategies
• Abstractions of Goal, Choice and BackTrack
• These abstractions hide
– Complex matching
– Variable value propagation and retraction
• Small footprint, easy to integrate as a subsystem
• Many reasoning tasks handled by systems built from
j-DREW
• Developed by Bruce Spencer
• Planned publication in implementations of logic
People
• Bruce Spencer
– RO and Adjunct Prof UNB CS
• Sandy Liu
– new RO, recent Masters from Acadia in E-Commerce
and Web Services
• Justin Hickey
– CS at NRC
• Ali Ghorbani
– Assoc Prof, UNB CS
• Harold Boley
– Senior researcher, DFKI, Germany
• Scott Buffett
– Guest worker and UNB PhD student
• Fang Wang and Keping Jia
– Guest Workers and UNB Masters students
Highlights
•
•
•
•
•
•
•
•
•
Sept 2001
– Bruce Spencer joins NRC as Research Officer
Nov-Dec 2001
– Harold Boley from DFKI (German Research Center for Artificial Intelligence) arrives as
Visiting Researcher
Dec 2001
– Said Tabet (Nisus Inc, MA) invited as Colloquium Speaker
Jan 2002
– Scott Buffett joins (Guest Worker and PhD Student in procurement planning systems)
Feb 2002
– Business Rules and Web Services Focus begins
– Dagstuhl Seminar on Rule Markup Techniques
– Invited talk at DFKI
Apr 2002
– Edited Special Issue of Computational Intelligence on Agent Technologies for E-Commerce
(11 papers, 200 pages) resubmissions from June 2001 workshop
– Keping Jia and Fang Wang join (Guest workers and Masters Students)
May 2002
– AI 2002 Conference in Calgary
• Co-edited proceedings, coordinated program and invited speakers
• Business Agents and the Semantic Web Workshop (Tabet, Boley, Ghorbani and Spencer)
June 2002
– E-learning Project begins with Mosaic Information Technologies
July 2002
– Sandy Liu joins from Acadia (Research Officer)