Posts Tagged ‘Web Ontology Language’

Introduction to Description Logic in Medicine


Description Logic based applications have been developed and successfully used in various domains. Many of those applications like KL-ONE [5], LOOM [6], and KRIS [7] have all used DL or some variant of DL to address real world problems. Description Logic has been recognised to have applications in medicine in areas of decision support, intelligent user interfaces, terminologies and natural language generation [8]. Though OWL has been only accepted as a W3C 1 recommendation in 2004, various DL variants have been used in the past to build and maintain medical terminologies. Snomed CT which is currently recognised as the international standard medical terminology reference uses DL to capture knowledge in the form of concept hierarchies and relationships. A variant of DL called KRSS is used to model and maintain the structure of the hierarchy in Snomed. OpenGALEN which is another medical terminology geared towards surgical procedures is based on another variant of DL which is GRAIL. KRSS and GRAIL are variants of DL and predecessors of the more popular Knowledge representation language, OWL. Before we discuss the advantages of using OWL-DL knowledge bases, it is useful to understand the general architecture of an OWL-DL ontology.

Features of DL Ontologies/ Terminologies

A formal ontology has a vocabulary that is defined in terms of ‘concepts’ and ‘roles’ which define the relationships between the concepts.

  • Primitive Concepts : These are the atomic building blocks of an ontology. These are often grouped together as hierarchies. E.g.: Arm, Forearm, Hand, Finger etc
  • Properties : These represent roles that are then used to define relationships between primitive concepts. E.g.: ‘has part ’ , ‘has action ’ etc
  • Defined Concepts : These are complex descriptions built using primitive concepts and properties.
  • Restrictions : These are property-concept pairs qualified by logical attributes like ‘some ’ and ‘only ’ . E.g.: ‘has part ’ some ‘Hand ’
  • Axioms: These are statements or assertions about named concepts in the ontology. For example the statement that a Concept A is related to a Concept B via a property ‘is part of ’ is an axiom. e.g. A ‘is part of ’ B
  • Reasoner: A service built into DL based ontologies, which allows ‘reasoning’ over an ontology. Reasoning usually consists of determining if a concept or description is a child of another (subsumption ) . It also checks axioms and descriptions in the ontology for logical consistency.
  • TBox : This is the terminology, i.e., the vocabulary which describes the domain. This is usually the set of all concepts and properties used in the knowledge base.
  • ABox: This is the set of all assertions about the named individuals in the domain described in terms of the vocabulary.

These basic building blocks are used to build complex concepts in the OWL-DL language.

The Web Ontology Language: OWL

The Web Ontology Language is an expressive formalism intended to model knowledge using concepts and relationships based on a formal syntax. It is designed for use by software agents that need to process the content of information. It differs from XML and XML Schema in that it offers a means to represent knowledge
and not a message format (which is the case with XML). The availability of tools to reason over an OWL ontology, allows knowledge represented in the ontology to be used for answering queries.

The OWL language has three species of increasing expressivity for use by specific communities or modellers: OWL Lite, OWL DL and OWL Full.

  1. OWL Lite supports building ontologies with simple constraint features and mainly to maintain classification hierarchies. Some of the features supported are restricted; eg. cardinality constraints can only have values of 0 or 1. OWL Lite is intended as a quick migration path for existing resources. Though there there are no constraints for building new ontologies in OWL Lite, most complicated ontologies require more expressive power offered by OWL DL or OWL Full.
  2. OWL DL allows the use of maximum expressiveness without losing computational tractability. OWL DL includes all OWL language constructs with some restrictions like type separation 2 . Since there is a correspondence between Description Logics (a variant of decidable first order logic), this species of OWL is called OWL DL. Since an OWL DL (and OWL Lite) uses only those features of the OWL language that allow complete and sound reasoning, it is ideal to support complex ontologies which can drive applications.
  3. OWL Full supports all features of the OWL language without any restrictions. For example, in OWL Full a class can be treated simultaneously as a collection of individuals and as an individual in its own right. However there is no guarantee that OWL Full ontologies are decidable. This makes it hard to use OWL Full to develop DL driven applications.

Since each OWL species is an extension of its predecessor, it can be said that:

  • Every legal OWL Lite ontology is a legal OWL DL ontology.
  • Every legal OWL DL ontology is a legal OWL Full ontology.

It is also worth remembering that of the three species of OWL, only OWL Lite and OWL DL are decidable.

References:

1. RA Miller, HE Pople, and JD Myers. Internist -i, an experimental computer- based diagnostic consultant for general internal medicine. New England Journal Of Medicine, 397:468–476, 1982.
2. E.H. Shortliffe. MYCIN: A Rule-Based Computer Program for Advising: Physicians Regarding Antimicrobial Therapy Selection. PhD thesis, Stanford University, 1974.
3. EH Shortliffe, A Carlisle Scott, Miriam B Bischoff, A Bruce Campbell, William van Melle, and Charlotte D Jacobs. An expert system for oncology protocol management. In BG Buchanan and EH Shortliffe, editors, Rule Based Expert Systems: The Mycin Experiments of the Stanford Heuristic Programming Project, pages 653–665. Addison-Wesley, Reading, Mass, 1981. Originally published in Proceedings of the Seventh International Conference on Artificial Intelligence (IJCAI-81) pp 876-881.
4. S. Bechhofer, F. van Harmelen, J. Hendler, I. Horrocks, D.L. McGuinness, P.F. Patel-Schneider, L.A. Stein, et al. OWL Web Ontology Language Reference. W3C Candidate Recommendation 18 August 2003. http://www. w3.org/TR/2003/CR-owl-ref-20030818, 2003.
5. RJ Brachman and JB Schmolze. An overview of the kl-one knowledge representation system. Cognitive Science, 9(2):171–216, 1985.
6. R MacGregor and R Bates. The Loom Knowledge Representation Language.
7. ISI reprint Series, University of Southern California Press, Los Angeles, 1987.
8. F. Baader and B. Hollunder. KRIS: Knowledge Representation and Inference System. ACM SIGART Bul letin, 2(3):8–14, 1991.
9. Alan L Rector. Description logics in medical informatics. In Franz Baader, Diego Calvanese, Deborah L McGuinness, Daniele Nardi, and Peter F Patel-Schneider, editors, The Description Logic Handbook: Theory, implementation and applications, pages 406–426. Cambridge University Press, Cambridge, England/ New York, 2002.
10. Franz Baader, Diego Calvanese, Deborah L McGinness, Daniele Nardi, and Peter F Patel-Schneider, editors. The Description Logic Handbook. Cambridge University Press, Cambridge, England, 2003.
11. J. Graunt. Natural and Political Observations Made upon the Bills of Mortality. London, UK: 1662; reprinted Baltimore. Md: The Johns Hopkins Press, 1939.
12. Alan L Rector. Clinical terminology: Why is it so hard? Methods of Information in Medicine, 38:239–252, 1999.