Ontologies

1
Ontologies

• EEL 5937 Multi Agent Systems
• Lecture 5, Jan 23th, 2003
• Lotzi Bölöni

2
Ontology editorProtégé-2000
3
Protégé-2000

• Developed at Stanford Medical Institute
• Java based ontology and knowledge base editor
• A tool which allows the user to
• construct a domain ontology
• customize knowledge-acquisition forms
• enter domain knowledge
• A platform which can be extended with graphical
  widgets for tables, diagrams, animation
  components to access other knowledge-based
  systems embedded applications
• A library which other applications can use to
  access and display knowledge bases.

4
Protégé-2000 as a library

• Public Java API allows procedural access to the
  knowledgebase
• Create, modify delete knowledgebases
• Create, modify, delete classes and slots.
• Create, modify, delete instances lt-- this is how
  you will use it in your projects.
• Also allows access to the graphical elements of
  Protégé (e.g. forms).

5
Protégé-2000 Backends

• Native format
• Clips inspired frame representation
• Plain text, Lisp like language
• Plugins for backends
• XML storage backend
• JDBC backend storing the knowledgebase in a
  relational database
• RDF backend
• Problems
• No support for concurrent access.
• Simplistic mapping to relational databases.
• Limited performance.

6
Protégé-2000 Plugins

• PAL Protégé Axiom Language
• A superset of first order logic.
• Can be used to express constraints about the
  knowledge base
• Can be used to make queries about the knowledge
  base
• JessTab integration with Jess
• Jess is a Java based implementation of the Clips
  expert system shell.
• PROMPT interactive ontology merging tool
• Multiple ontologies problems of mapping,
  refactoring etc.
• XML
• allows importing XML documents into
  Protégé-2000, creating a set of classes and
  instances dynamically.
• Allows the saving of a Protégé-2000 database into
  XML.

7
Project for a massive ontologyCyc
8
Cyc

• Founded by Doug Lenat (former professor at
  Carnegie Mellon and Stanford).
• Cyc software has been under development since
  1984, founded mostly by government contracts.
• We stand on the threshold of success
• A knowledgebase of human common sense
• The Cyc knowledge base has an ontology of over
  100,000 atomic terms axiomatized by a set of over
  1 million handcrafted assertions stated in an
  n-th order predicate calculus employing over
  10,000 predicates which are themselves first
  class terms in the knowledgebase.

9
Cyc (contd)

• The Cyc inference engine includes general theorem
  provers but maintains efficiency by relying on a
  suite of over 500 heuristic level modules
• The knowledge base is divided into locally
  consistent contexts called micro-theories.
• Each micro-theory contains
• Content (a body of assertions)
• Assumptions shared by those assertions
• The variant of the predicate calculus used to
  represent the assertions is a language called
  CycL.
• Lisp syntax
• Interesting fact they do not use neither fuzzy
  logic, nor probabilities

10
OpenCyc

• http//www.opencyc.org
• open source version of the Cyc technology
• Not quite completely open, knowledgebase is not
  listable, some tools are binary
• Parts of the ontology (e.g. upper ontology) are
  available.
• Version 0.7beta released on December 17, 2002
• It is implemented as a web-service. You start it
  as a process and browse it through your browser.
• You are encouraged to use it in your projects!
  (if you can find a use for it).

11
Evaluation of the Cyc technology

• Ambitious, huge effort. Painstaking, slow
  knowledge engineering process.
• Probably, they have envisioned that the
  autonomous learning threshold will be reached
  much sooner.
• Questions
• Is the model correct?
• Were the original assumptions correct?
• If they were not, it would be very expensive to
  discover it now
• Researchers outside have a mixed opinion.
• Using it is not trivial, exactly because of the
  size of the DB.

12
DMTF / CIM
13
DMTF / CIM

• Distributed management task force an industry
  organization to promote the unification and
  adoption of management standards for desktop,
  enterprise and internet environments.
• http//www.dmtf.org
• CIM Common Information Model
• A model for describing overall management
  information in a network / enterprise
  environment
• Basically, an ontology under a different name
• CIM Specification
• The language and style of the descriptions.
• How to describe common patterns relations,
  associations etc.
• Mapping to other management models (e.g. SNMP)
• Most of the CIM specifications are visually
  presented in UML.
• They are formally expressed in MOF (Managed
  Object Format) plain text, and quite similar to
  Clips.

14
CIM Schemas

• The actual ontologies
• Core Schema is an information model that captures
  notions that are applicable to all areas of
  management
• Common Schemas are information models that
  capture notions that are common to particular
  management areas, but independent of a particular
  technology or implementation. The common areas
  are systems, devices, networks, applications,
  metrics, databases, the physical environment,
  event definition and handling, management of a
  CIM infrastructure (the Interoperability Model),
  users and security, policy and trouble ticketing/
  knowledge exchange (the Support Model).
• Extension Schemas represent organizational or
  vendor-specific extensions of the Common Schema.
  These schemas can be specific to environments,
  such as operating systems (for example, Unix or
  Windows).