1. AMIA Workshop WP06 Ontological Foundations of Biomedical Terminology Systems
Organized by AMIA WG Formal (Bio)Medical Knowledge Representation
Washington, DC, 10/22/2005

2. History
2002: Ronald Cornet / Stefan Schulz: Idea of foundation of a SIG on Ontologies / Knowledge Representation
2003: SIG “Formal (Bio)Medical Knowledge Representation. Approval by AMIA, Kick-Off Meeting at AMIA KR-SIG Mailing List
2004: KR-MED 2004 in Whistler (Canada)
Panel at MEDINFO
Upgrade SIG -> WG
2005: Intensive discussions on basic notions of terminology / ontology building, draft of a AMIA paper “Six questions on ontology building” (not accepted)
AMIA Workshop submission. Purpose of workshop: To stimulate and intensify the discussion on formal and philosophical foundations of biomedical ontologies.
Plan for 2006: Write a WG position paper -> JAMIA
Tuesday 5:15 p.m. Business meeting of the WG: Thoroughbred room, Concourse level.

3. Workshop WP06 Ontological Foundations of Biomedical Terminology Systems
Topic
Presenter
Duration (min)
Introduction
Stefan Schulz 5
1. Resolving ambiguities in current uses of the term "concept"
Barry Smith
(Q&A)
2. Do divergences in philosophical thinking (Realism, Conceptualism, Nominalism) have implications the concrete tasks of ontology engineering in the biomedical domain?
Stefan Schulz
(Q&A)
3. Should the term 'ontology' be used in a broad sense which does not exclude any of those artifacts commonly referred to as 'biomedical vocabularies' in the medical informatics community?
Ronald Cornet
(Q&A)
4. Discussion on the basis of the three talks
Led by Ira Kalet 10
Break 5
5. How can ontologies, which deal with relations between universals, be made interoperable with the electronic health data, which deal with instances of such universals?
tbd.
(Q&A)
6. Facilitation of ontology creation by human curators using statistical and lexical approaches
Olivier Bodenreider
(Q&A)
7. Difference between pre and post-coordinated ontologies and its implications on single and multiple inheritance. Should the design of formal ontologies be guided by concrete recommendations, e.g., single inheritance, avoidance of negative classes, pairwise disjoint siblings, jointly exhaustive siblings?
Anand Kumar
(Q&A)
Wrap-up
Ronald Cornet 10

4. Realism, Conceptualism, Nominalism and their Impact on Biomedical Terminologies / Ontologies

5. Stefan Schulz Martin Boeker Freiburg University Hospital (Germany)
Realism, Conceptualism, Nominalism and their Impact on Biomedical Terminologies / Ontologies
as naïvely understood by non-philosophers…
Stefan Schulz
Martin Boeker
Freiburg University Hospital (Germany)

6. What do the nodes in biomedical terminology systems stand for?
types
names
sets
universals
sorts
categories
descriptors
entities
synsets
Biomedical Terminology construction and maintenance is one of the most prominent fields in Medical Informatics and has a major impact on medical documentation and health services management. However, there is a range of seemingly disparate views represented by experts in the field. A symptomatic example for this is the difficulty to find not only commonly agreed words but even commonly agreed notions for what the nodes in biomedical terminology systems stand for. The assortment of names is impressive. Each of these names reflects different traditions and ways of thinking about closely related, albeit not equivalent notions.
classes
terms
properties
descriptors
concepts

7. Terminological Confusion
Different scientific traditions (Logics, Computer Science, Library and Information Science, Cognitive Science, Linguistics,…)
Different philosophical schools of thinking: (Platonism, Aristotelian Realism, Conceptualism, Relativism, Idealism, Postmodernism, Constructivism, Nominalism, Tropism,…)
Is there a common ground which suits the needs of Biomedical terminology / ontology construction ?
The terminological confusion is due to different scientific traditions and different schools of philosophical thinking. Our question is whether there is a common ground which suits the needs of Biomedical terminology / ontology construction.

8. Mind Independent Reality
I here try to sketch a framework of the different schools of thinking which have a long tradition among philosophers but are only vaguely rooted in the minds of most other people.
The common ground I propose is that there exists a mind independent reality of individual objects. Without this understanding, it would be difficult perform research. Even who casts doubts on such a reality, commonly takes it as a handy working hypothesis in order to interact with it and to communicate about it.
The recent controversy in Medical Informatics about the best philosophical approach to the “nodes” in terminological systems is due to different schools of thinking which have a long tradition among philosophers but are only vaguely rooted in the mind of most other people. As long as there is no or little awareness of these distinctions, the authors of biomedical terminologies and classification systems often base their systems on implicit and subconscious assumptions and end up with promiscuous systems. If they were trained in philosophical ontology (what cannot be expected on a broad scale) they would certainly streamline their thinking along one of the traded lines and, as a consequence, they rarely would reach an agreement. Since the rationale of medical terminology is to attain semantic interoperability between humans and machine agents, we have to identify the common ground.
One common ground is the acceptance of a mind independent reality. Without this understanding, it would be difficult to be a researcher. Even who has doubts about such a reality, commonly takes it as a handy working hypothesis in order to interact with it and to communicate about it. Once we accept such a reality we
Mind Independent Reality
Individual Entities (Instances, Particulars)

9. Mind Independent Reality
Schools of Thinking
Realist
Realist
Conceptualist
Nominalist
Universals
Universals
Concepts
Names
Categories
There are three main schools of philosophical thinking all of which have a certain influence on the practice of biomedical terminologies.
Firstly, Realism postulates that there exist universal properties, as a kind of templates of reality. These properties exist independently of human cognition.
As intelligent beings we can investigate these properties as instantiated in particular objects. For example, the universal human being is instantiated by
every human, the universal “chair” is instantiated by every chair and so on.
Mind Independent Reality
Individual Entities (Instances, Particulars)

10. Mind Independent Reality
Schools of Thinking
Realist
Realist
Conceptualist
Conceptualist
Nominalist
Universals
Universals
Concepts
Concepts
Names
Categories
An alternative view, called conceptualism propagates that general or universal entities are concepts, i.e. entities of human cognition. At least the more prominent proponents of this theory however, accept the existence of mind independent individual entities. However they insist that the principles of structuring these entities cannot be detached from cognition.
Mind Independent Reality
Individual Entities (Instances, Particulars)

11. Realist Conceptualist Nominalist Universals Concepts Names
Schools of Thinking
Realist
Conceptualist
Nominalist
Universals
Concepts
Names
Categories
trilateral
rectangle
square
circle
triangle
square
circle
Reference
to Reality
Finally, nominalists, deny that there are universals beyond linguistic predicates.
Mind Independent Reality
Individual Entities (Instances, Particulars)

12. Ordering / Organizing Reality
Two approaches:
Delimitation of individual entities: identifying aggregation as mereological sums of atomic individuals
Classification: ordering individuals: assignment to classes
What’s the implication of these different schools for the two major task for terminologists,
namely first, the need to delimit individual entities, it means to determine the boundaries of entities,
e.g. the boundary of brain regions or the boundary of the liver. and secondly the classification
of entities.
Individual Entities (Instances, Particulars)

13. Delimitation of Individual Entities
Problem: which chunks of matter / sum of atoms are considered valid composite entities?
Realist: composition / delimitation of individual entities as pattern of being
Conceptualist / nominalist: composition / delimitation of individual entities as result of human cognition or decision (“fiat”)
Individual Entitity

14. Delimitation of Individual Entities
Problem: which chunks of matter / sum of atoms are considered valid composite entities?
Realist: composition / delimitation of individual entities as pattern of being
Conceptualist / nominalist: composition / delimitation of individual entities as result of human cognition or decision (“fiat”)
Individual Entitity

15. Delimitation of Individual Entities
Problem: which chunks of matter / sum of atoms are considered valid composite entities?
Realist: composition / delimitation of individual entities as pattern of being
Conceptualist / nominalist: composition / delimitation of individual entities as result of human cognition or decision (“fiat”)
Individual Entitity

16. Delimitation of Individual Entities
Problem: which chunks of matter / sum of atoms are considered valid composite entities?
Realist: composition / delimitation of individual entities as pattern of being
Conceptualist / nominalist: composition / delimitation of individual entities as result of human cognition or decision (“fiat”)
Individual Entitity

17. Delimitation of Classes
Problem: which individual entities belong to the same class?
Realist: class membership as invariant pattern of being (instantiation of universals)
Conceptualist: class membership as extension of a concept due to perceived similarities
Nominalist: given name as criterion for class membership (strict N.: denial of classes)

18. Delimitation of Classes
Problem: which individual entities belong to the same class?
Realist: class membership as invariant pattern of being (instantiation of universals)
Conceptualist: class membership as extension of a concept due to perceived similarities
Nominalist: given name as criterion for class membership (strict N.: denial of classes)

19. Delimitation of Classes
Problem: which individual entities belong to the same class?
Realist: class membership as invariant pattern of being (instantiation of universals)
Conceptualist: class membership as extension of a concept due to perceived similarities
Nominalist: given name as criterion for class membership (strict N.: denial of classes)

20. Theoretical Upper Bound of number of individuals and classes
n atoms in reality: m = 2n -1 possible entities (arbitrary aggregations), candidates for individuals / particulars
entities: 2m -1 = different classes: possible extensions of universals, concepts, or names
2n -1
99, %
of these individuals and classes are completely irrelevant

21. Which classes are relevant…
… and correspond to the extension of universals, concepts or names ?

22. Realist Conceptualist Nominalist Universals Concepts Names
Schools of Thinking
Realist
Conceptualist
Nominalist
Universals
Concepts
Names
Categories
trilateral
rectangle
square
circle
triangle
square
circle
Reference
to Reality
2 objects {} a b ab
3 objects {} a b c ab bc ac abc
4 objects {} a b c d ab ac ad bc bd cd abc abd acd bcd abcd
n atoms in reality: 2^n -1 possible composite entities
m entities: 2^m -1 different classes
Classes:
Mind Independent Reality
Individual Entities (Instances, Particulars)

23. Realism: Classes as extensions of Universals:
All universals refer to non-empty (at some moment ) classes
New classes can be created by common set theoretical operations, e.g. disjunction, negation, complement.
Many of those classes do not correspond to a universal, e.g. “invertebrate”, “wound infection NEC”, … but such classes are nevertheless useful, e.g. in order to guarantee disjoint clinical classifications

24. Conceptualism: Classes as extensions of Concepts:
Unlike universals, concepts do not necessarily imply the extension to classes in reality (“retinal transplant”, “yin deficiency”, “missing digit”, “prevented pregnancy”)
Concepts as entities of thought may mix ontological and epistemological aspects: “Cirrhosis of liver without mention of alcohol”
Concepts as mind constructs may be oriented to prototypes, their extension exhibits large inter-individual variation (deviation from prototype)

25. Nominalism: Classes as extensions of natural language expressions
Classes are built in an ad hoc fashion from linguistic predicates.
Examples:
People in Room Monroe West at 8pm on October 22, 2005
“Extraction of Foreign Body from Stomach by Incision”

26. Universals and Continua
The realist approach is most striking when applied to individuals which show crisp “bona fide” distinctions and which are easy to classify, especially atoms and molecules
It runs into problems where the properties individuals exhibit lie on a continuum, which is the case with many biological objects.
property “epithelial”
property “neural”
neural cell
epithelial cell

27. Universals and Continua
The realist approach is most striking when applied to individuals which show crisp “bona fide” distinctions and which are easy to classify, especially atoms and molecules
It runs into problems where the properties individuals exhibit lie on a continuum, which is the case with many biological objects.
Universal A
Universal B
Universal C
neural cell
epithelial cell

28. Universals and Continua
The realist approach is most striking when applied to individuals which show crisp “bona fide” distinctions and which are easy to classify, especially atoms and molecules
It runs into problems where the properties individuals exhibit lie on a continuum, which is the case with many biological objects.
Universal V
Universal W
Universal X
Universal Y
Universal Z
neural cell
epithelial cell

29. Conclusions
Most terminology builders and users have a implicitly promiscuous attitude towards philosophical categories
Impossible to reach agreement on overly purist approaches
There may be reasons for hybrid use of universals, concepts, and names in biomedical terminologies
Talking about “classes of individuals” may be the best common denominator which allow bridging between different standpoints.
Nevertheless, awareness of philosophical / logical background should be raised among terminology builders and users

30. Nominalism: Classes as extensions of natural language expressions
Classes are built in an ad hoc fashion from linguistic predicates. Example: “Extraction of Foreign Body from Stomach by Incision”
RemovalOfForeignBodyFromDigestiveSystem AND
RemovalOfForeignBodyFromStomach AND
IncisionOfStomach AND
 has-part.( Method.RemovalAction AND
 DirectMorphology.ForeignBody) AND
 has-part.( Method.IncisionAction AND
 ProcedureSite.stomachStructure)