Rulebase Integration for eCollaboration

1
Rulebase Integration for eCollaboration
11th International Workshop on Telework,
Fredericton, NB, Canada

• Duong Dai Doan1 , Harold Boley2, Thuy Thi Thu Le1
  and Virendrakumar C. Bhavsar1

1Faculty of Computer Science, University of New
Brunswick,Fredericton, New Brunswick,
CanadaDuong_Dai.Doan, Thuy_Thi_Thu.Le,
bhavsar_at_unb.ca 2Institute for Information
Technology e-Business, NRC, Fredericton, New
Brunswick, Canada harold.boley_at_nrc-cnrc.gc.ca
August 28, 2006
2
Agenda

• Motivation
• Classification of Rulebase Integration Approaches
• Proposed Rulebase Integration Framework for
  Interoperation and Interchange
• Homomorphisms Semantics-Preserving
  Transformations
• Conclusion

3
Motivation

• Foundational work in semantic information
  integration central to cluster of Semantic Web
  projects at UNB and NRC Fredericton
• eBusiness (Weighted-Tree Similarity Algorithm)
• eLearning (RuleML)
• eCollaboration (FOAF)
• Define the objects and vocabularies of
  eCollaboration (e.g., merchandise, services) by
  rules
• Rules differ, syntactically, semantically, and
  pragmatically, between (groups of) participants
  of Web-based collaborations
• Rulebase integration techniques needed

4
Motivation

• The goal of Semantic Web community
• Semantic information sharing and reuse
• Underlying many Semantic Web applications
• Rule-based systems

5
Previous work

• Integration of databases
• Well-studied and long-standing challenge in the
  Database Community TSIMMIS, HERMES,LoPiX, HERA,
  and MIX
• A rule Head ? Body
• Head conclusion, Body condition
• When the body is empty, a rule is called a fact

discount(Customer, Product, "7.5 percent")-
premium(Customer),
regular(Product).

• (Relational) Database integration only deals with
  such kind of facts
• Database integration is regarded as special case
  of Rulebase integration XSIS

6
Classification of Rulebase Integration

• Heterogeneous rulebases can be serialized in
  different languages and use various fragments of
  expressiveness
• Classification is based on two main dimensions
• Language surface syntax
• Expressive(ness) fragment

7
Classification of Rulebase Integration
7
8
Integration of Rulebases Having the Same
Expressiveness

• Different rulebase providers have different
  perspectives on a collaboration
• Incoming rulebases are often heterogeneous
• Reconciliation conflicts between various
  rulebases is the key issue of rulebase
  integration
• Conflicts are classified into four main types
  (earlier slide)

9
Same Expressiveness - Synonyms

• Two terms (i.e., relations in rules) refer to the
  same object synonyms

RB1 merchandise(X) - provider(Y,X),
warehouse(Y,Z). provider("Compact Corp.",
"Printer"). warehouse("Compact Corp.","Boston").
RB2 item(X) - supplier(Y,X), store(Y,Z). supplier
("Compact Corp.", "Printer"). store("Compact
Corp.", "Boston").
Solution term dictionary provided for
one-to-one transformation between relation names
10
Relation subsumption

• P and P are on the same path in a relation
  subsumption hierarchy

Example P  merchandise  P product
subClassOf(merchandise, product)
subClassOf(P,P') is represented by P'(X) -
P(X). Example product(X) - merchandise(X).
11
Different Expressiveness

• Different rulebases can be formulated with
  different levels of semantics

• Gain Datalog gt Horn logic Loss Horn
  logic gt Datalog
• More expressive rulebase can be split for maximum
  interchange

12
(No Transcript)
13
Rulebase Interoperation vs. Interchange

• Rulebase interoperation performed on distributed
  (autonomous) rulebases
• Query transformation
• Distributed querying
• Answer composition
• Rulebase interchange
• Transforms heterogeneous rulebases into canonical
  form
• Using transformation rules, which themselves are
  interchangeable
• Information may be preserved or lost through the
  transformation
• This transformation can thus be total or partial
• Supports not only uniform querying but also
  processing

14
Homomorphisms Semantics-Preserving
Transformations

• Transform a rulebase encoded in a language to
  another language
• Information is preserved during transformation
• Some information may be lost lossy
  transformation

trans-1(infer(trans(XDD))) infer(XDD)
15
Conclusion

• Define the objects and vocabularies of
  eCollaboration by rules (including taxonomies)
• Classification of rulebase integration
• Resolution of conflicts
• Unified framework for rulebase integration
• Interoperation approaches
• Interchange approaches
• Homomorphisms for preserving inferential
  semantics of rulebases on transformation
• Rulebase interchange experiments with XSLT
  stylesheets
• XDD and RuleML 0.9
• RFML and (Functional) RuleML 0.91

16
Thanks !
17
References

• XML Schema Integration System (XSIS)
• http//people.unb.ca/b89ct
• RuleML homepage
• http//www.ruleml.org
• RuleML FOAF-A Use Case for Web-based Social
  Networking
• http//www.ruleml.org/usecases/foaf/
• AgentMatcher
• http//www.cs.unb.ca/agentmatcher/