Transcript WG2-SEL-013-Common-Logic-&-Ontologies

Proposed NWI
KIF/CG --> Common Logic Standard

A working group was recently formed from the KIF working group.
John Sowa is the only CG representative so far.
– Stanford March 2002:



Agreed on the name: Common Logic Standard for First Order Logic
Cris Menzel (U.Texas, Austin Philosophy dept) agreed to maintain a web
site, now called
Conceptual Graphs and KIF to be reference implementations.
– KIF maintained by KIF working group, including Mike Genesereth who was
the contact for requirements during its development.
– CG draft standard now available at a site maintained by John Sowa.
Requirements are discussed in CGlist and at standards meetings at the annnual
ICCS conference (2002, Bulgaria). Previously held in Canada, Germany, US,
France.
– CG and KIF communities both contain competing vendors and academics.

Other ISO logic standards to be described in relationship to the
Common Logic
– Prolog Logic Programming Language (Mike Gruninger)
– Object Constraint Language (under UML)
– Z Formal Specification Language
Common Logic Standard

Will support:
– Institutionalization of KIF and CG
Periodic versions instead of unscheduled ad-hoc changes will
encourage tools development
 Researchers and Developers can make requests of a standards body
rather than of individual caretakers, who may have business or
professional interests to protect.
 Contract requirements can specify differences from and extensions to
the Common Logic standard used by vendor specific languages such
as Cycorp’s Cyc-L, Ontos’ , XSB, Haley Eclipse, NASA CLIPS…)
– Specification of axiomatic ontologies
– Selection of a knowledge representation language for a purpose
 Expressivity/ computablility tradeoffs
– Translation between formal languages
 For changing vendors, user interfaces, performance problems

Common Logic in support of
Axiomatized Ontologies

Ontology working definition (Gruber)
– Formal (to support automated inference)

explicit specification (concepts and relations are given terms and definitions)
– of a shared conceptualization (abstract model relating people’s thinking to things in
the world (usually in a domain)
–
A community with a shared conceptualization may have both a common natural
language and domain ( Northumberland sheep herders ) or a common domain and
different natural language ( Mathematics, logic?).
– (Gruninger and Lee) Ontology must capture the intended semantics of the users
teminology.
• Not only ontology must capture meaning, but any component of an
information system which has a human interpreter.


Ontology structure: hierarchy of concepts or classes and their
relationships
Purpose: Capture world knowledge or (at least) domain understanding
Ontologies and Medical Informatics

Application example: creation of structured information from medical texts:
– Discharge Summaries
– Requests
– Patient Descriptions


Purpose: enable queries against the resulting database
See Marie Rassineoux’s work for Geneva Hospital
– Application: creation of structured information from medical texts:

Discharge Summaries, Requests, Patient Descriptions
– Purpose: to enable queries against the resulting database
– Concept classes for NLP for Gastro-Intestinal Surgery

Root
–
–
–
–
–
–
Actor
Event
Attribute
Value
Temporal
Relationship
Cyc Upper Level
THING

– Individual Object
Event
Stuff
Intangible
Intangible Object
Represented Thing
Collection
Lattice shaped
Relationship is an intangible thing, with slots.
Axioms expressed in Cyc-L, full first order logic, with second order extensions.
Supporting Ontologies
can connect an Information System to the world
of the end user.

Knowledge representations richer than SQL and more
accessible to the end user permit:
– Formation of complex queries
– Construction of templates for pattern matching and data mining in dense
information environments
– Collaborative ontology maintenance
– User participation in lexicon maintenance
– Formation of queries against heterogenous databases, allowing their concepts
to be mapped using more general ones.
Upper Level Ontology:
Standardize? Or Register & Map

Influences on Upper Levels
– Philosophy
– Linguistics
– Pragmatics
– Business Plans
– Academic School of Thought
– Domain Ontologies



Intensional approach (SUMO) needs agreement on theory
Extensional approach (Gruber, FCA) mine data models, documents, thesauri
Registration of Ontologies
– 11179 Part 2 as a classification scheme
– An ontology may be used to generate a data model or a value domain, either in whole or
in part. It may evolve more or less continuously, with the data model being regenerated
less often. The linkage is not simple.
– DAML + OIL expect ontology mapping services on the web
Related Standards Efforts

CSMF ISO/IEC 14481
withdrawn
FCD 12/1/98 -Proposed to be
– Universe of Discourse for an enterprise and its users
– Formal description for information systems design, interchange and
integration, defined using typed predicate logic
– Entities, properties , relationships, constraints, inheritance mechanisms,
rules, events, functions, processes.
– Issues overlap those of 11179 P.2

ISO TR 9007 SC32, 1987
– Schema is independent of technology or platform.
– Exchange of meaning requires agreeement

DAML
– Proposed for description of ontologies governing semantics of web page
markup. Uses Description Logics and ontology mapping services.