Presentation is loading. Please wait.

Presentation is loading. Please wait.

Representing the UMLS Semantic Network using OWL

Similar presentations


Presentation on theme: "Representing the UMLS Semantic Network using OWL"— Presentation transcript:

1 Representing the UMLS Semantic Network using OWL
Vipul Kashyap1 and Alex Borgida2 1 LHCNBC, National Library of Medicine, 8600 Rockville Pike, Bethesda, MD 20894 2 Department of Computer Science, Rutgers University, New Brunswick, NJ 08903 Seminar Prinzipien des Ontological Engineering Leipzig, Kristin Lippoldt

2 Outline The UMLS Semantic Network (SN) Representation of SN using OWL
Multiple interpretations of „link“ Evaluation of the interpretation variants Methodology for choosing the „right“ representation variant (first steps)

3 The UMLS Semantic Network
nodes = semantic types links = semantic relationships two high level is-a hierarchies Entity, Event is-a hierarchie of relationships physically_related_to, spatially_related_to, temporally_related_to, functionally_related_to, conceptually_related_to functionally_related_to affects is-a manages is-a

4 The UMLS Semantic Network (excerpt)

5 OWL Web Ontology Language Based on DAML+OIL
Description of classes, properties (e.g. relations between classes (e.g. disjointness), cardinality (e.g. "exactly one")) Sublanguages: OWL Lite (lower formal complexity than OWL DL, only cardinality values of 0 or 1) OWL DL (maximum expressiveness, computational completeness ) OWL Full (maximum expressiveness, syntactic freedom of RDF with no computational guarantees) SN semantic type = DL primitive concept = OWL class SN relationship = DL primitive role = OWL object property

6 Description Logic - OWL
Bacterium ODER Virus <owl:Class> <owl:unionOf rdf:parseType=“Collection”> <owl:Class rdf:about=“#Bacterium”/> <owl:Class rdf:about=“#Virus”/> </owl:unionOf> </owl:Class> SN semantic type = DL primitive concept = OWL class SN relationship = DL primitive role = OWL object property

7 Representation of SN using OWL
Semantic Types  OWL classes Fungus  Organism Virus  Organism Semantic Relationships  OWL properties part_of  physically_related_to affects  functionally_related_to Properties of Semantic Network Relationships Asymmetric relationships has_part ≡ part_of Symmetric relationships adjacent_to ≡ adjacent_to - Pilz ist Unterklasse von Organismus mit OWL properties kann die Relationshierarchie umgesetzt werden adjacent = angrenzend

8 Semantics of a „link“ in the UMLS SN
Bacteria causes Infection Two operators  and : (causes) = { x  Bacteria  (y)(y  Infection  causes(x,y)) } DL notation: (causes) ≡ causes.T (causes) = { y  Infection  (x)(x  Bacteria  causes(x,y)) } DL notation: (causes) ≡ causes.T (causes) ≡ causes.T  alle verursachenden Top-Konzepte (linke Seite) (causes) ≡ causes.T  alle verursachten Top-Konzepte (rechte Seite)

9 Interpretation 1: / equals
axioms: causes.T ≡ Bacteria, causes.T ≡ Infection All Bacteria have to “cause” and all Infections have to “be-caused” (no others can participate in “causes”) b1 i1 b2 i2 b3 i3 b4

10 Interpretation 2: / subsumed
axioms: causes.T  Bacteria, causes.T  Infection Not all bacteria need to “cause” not all infections have to “be-caused” (However no others can participate) i1 b2 i2 die Entitäten, die verursachen ist eine Subklasse der Klasse der Bacteria die Entitäten, die verursacht werden ist eine Subklasse der Klasse der Infection b3 i3 b4

11 Interpretation 3: / subsumes
axioms: Bacteria  causes.T, Infection  causes.T All bacterias have to “cause” and all infections have to “be-caused”, but A bacteria can cause a “non-infection” as well! A “non-bacteria” can cause an infection as well! y1 i1 Bacteria ist eine Subklasse der Entitäten, die verursachen Infection ist eine Subklasse der Entitäten, die verursacht werden b2 i2 b3 i3 b4 x1

12 Interpretation 4: All/Some
axiom: Bacteria  causes.Infection All bacteria must “cause” some infection, but A bacteria can cause a “non-infection” as well! A “non-bacteria” can cause an infection as well! y1 i1 Bacteria ist eine Subklasse der Entitäten, die verursachen Infection ist eine Subklasse der Entitäten, die verursacht werden b2 i2 b3 i3 b4 x1

13 Interpretation 5: All/Only
axiom: Bacteria  causes.Infection All bacteria, if they “cause”, can cause only infections, but Not all bacteria have to participate in the “causes” relationship A non-bacteria can still cause an infection! y1 i1 A non-bacteria can still cause a non-infection! b2 i2 b3 i3 b4

14 Interpretation 6: All/Each
axiom: Bacteria  causes.Infection Similar to a cross product, but A bacteria can still cause a non-infection! i1 Bacteria   Infection.causes jedes Bakterium verursacht alle Instanzen der Klasse Infektion b2 i2 b3 i3 b4 x1

15 Interpretation 7: Some/Some
axiom:  1 (Bacteria  causes.Infection) There is at least one bacteria that “causes” at least one infection, but A bacteria can still cause a non-infection! A non-bacteria can still cause an infection! y1 i1 Durchschnitt b2 i2 b3 i3 b4 x1

16 Interpretation 8: Some/Each
axiom:  1 (Bacteria  causes.Infection) There is at least one bacteria that “causes” all infections, but A bacteria can still cause a non-infection! A non-bacteria can still cause an infection! y1 i1 Schnittmenge zwischen Bacteria und allen Dingen, die Infektionen verursachen b2 i2 b3 i3 b4 x1

17 Summary of Interpretations
equals: causes.T ≡ Bacteria, causes.T ≡ Infection subsumed: causes.T  Bacteria, causes.T  Infection subsumes: Bacteria  causes.T, Infection  causes.T all/some: Bacteria  causes.Infection all/only: Bacteria  causes.Infection all/each: Bacteria  causes.Infection some/some:  1 (Bacteria  causes.Infection) some/all:  1 (Bacteria  causes.Infection)

18  and  Inheritance  inheritance P(A,B) C  A P(C,B)
 inheritance P(A,B) D  B P(A,D) Example: process_of(BiologicFunction,Organism) C = PhysiologicFunction D = Animal equals: no support of inheritance , A ≡ C subsumed: no support of inheritance alle verursachenden Entitäten sind äquivalent zu A und ebenfalls zu C, also muss A äquivalent zu C sein C ist subconcept von A, alle verursachenden Entitäten sind subconcept von A, daraus folgt aber nicht, dass alle verursachenden Entitäten ebenfalls subconcept von C sind A C process_of.T

19  and  Inheritance process_of.B subsumes: supports both
process_of.T process_of-.T subsumes: supports both all/some: supports  inheritance, but not  inheritance all/only: supports  inheritance, but not  inheritance A B C D process_of.B process_of-.D A B 4) C = Staphylokokken, aber D = Streptokokkeninfektion 5 ist genauso wie 4 C D process_of.B A C

20  and  Inheritance process_of. D process_of. B
all/each: supports both some/some: no support of inheritance some/all: doesn’t supports  inheritance, but  inheritance process_of. B A C

21 Blocking of Inheritance
Example: Process_of(BiologicFunction,Organism) Process_of(MentalProcess,Plant) Modifying axioms: subsumes: P(A,B) C1  A and D1  B A  C1  (P) and B  D1  (P)

22 Polymorphic Relations
Ergebnis Interpretation Encoding / Inheritance Inheritance Blocking Polymorphic Relations / equals (P)  A (P)  B No/No N/A No / subsumed (P)  A (P)  B Missed model / subsumes A  (P) B  (P) Yes/Yes Exceptions + compensation Unintended model all / some A  P.B Yes/No Exception in axiom ok all / only A  P.B Modification some / some 1(A  P.B) some / all 1(AP.B) No/Yes all / each A  P.B

23 Methodologie für die Kodierung von Wissen im Semantic Web
Wahl der Kodierung Unterstützung von Inferenz Unterstützung der intendierten Anwendung Nachvollziehbares Domänenmodell Repräsentation in der Ontologiesprache

24 Unterstützung von Inferenzen
Welche Kodierung unterstützt Inferenz? All/each und subsumes Unterstützt die Kodierung nicht-intendierte Inferenzen? Some/some unterstützt Aufwärts-Vererbung von Links Kann etwas aus der Abwesenheit eines Links geschlussfolgert werden? A  P. B verbietet nicht, dass A in Relation zu B steht

25 Unterstützung der intendierten Anwendung
Ist es wichtig Inkonsistenzen zu erkennen? Was sind Inkonsistenzen? Wird die Kodierung diese Inkonsistenzen erkennen?

26 Nachvollziehbarkeit des Domänenmodells
Konzepte sind Kollektionen von Instanzen Causes(Bacteria,Infection) Was ist die intuitive Kodierung? All/some and all/only wird von medizinischen Ontologien genutzt All/each und some/some wurden abgelehnt Gibt es alternative Interpretationen? Aber: all/each erfüllt alle UMLS SN Anforderungen

27 Repräsentation in der Ontologiesprache
Grenzen von OWL Negation und Disjunktion von Rollen Kardinalität von Konzepten Kann man weniger „teure“ Konstrukte verwenden? Ressourcen fließen in die Komplexität der DL Operatoren

28 Conclusions and Future Work
Experiences in representing a real world “ontology”, the UMLS Semantic Network Has been used very successfully Requirements: / inheritance, inheritance blocking, polymorphic relationships Presented multiple interpretations and encodings and evaluated their support for the UMLS Semantic Network requirements Ontology developers and encoders on the Semantic Web might encounter similar requirements and possible encodings Identified criteria for choosing between the various encodings First steps towards a methodology which might be useful to ontology developers Ongoing and Future Work Semantic Vocabulary Interoperation Project Use of OWL, RDF for improvement in Medical Information Retrieval


Download ppt "Representing the UMLS Semantic Network using OWL"

Similar presentations


Ads by Google