Summary Issues and Suggestions Workshop on The Future of the UMLS Semantic Network NLM, April 8, 2005 Olivier Bodenreider Olivier Bodenreider Lister Hill National Center Lister Hill National Center for Biomedical Communications for Biomedical Communications Bethesda, Maryland - USA Bethesda, Maryland - USA
22
Embed
Summary Issues and Suggestions Workshop on The Future of the UMLS Semantic Network NLM, April 8, 2005 Olivier Bodenreider Lister Hill National Center for.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
SummaryIssues and Suggestions
Workshop onThe Future of the UMLS Semantic Network
NLM, April 8, 2005
Olivier BodenreiderOlivier Bodenreider
Lister Hill National CenterLister Hill National Centerfor Biomedical Communicationsfor Biomedical CommunicationsBethesda, Maryland - USABethesda, Maryland - USA
Issues
3 Lister Hill National Center for Biomedical CommunicationsLister Hill National Center for Biomedical Communications
UMLS Semantic NetworkUMLS Semantic Network
Necessary complement to the MetathesaurusNecessary complement to the Metathesaurus Provides direct categorization to conceptsProvides direct categorization to concepts
(some of which would be orphans otherwise)(some of which would be orphans otherwise)
Best used in conjunction with the MetathesaurusBest used in conjunction with the Metathesaurus Used forUsed for
Natural Language ProcessingNatural Language Processing Information retrievalInformation retrieval Knowledge discoveryKnowledge discovery
Essentially stableEssentially stable
4 Lister Hill National Center for Biomedical CommunicationsLister Hill National Center for Biomedical Communications
Semantic typesSemantic types
Purposely limited to a small number of categoriesPurposely limited to a small number of categories Purposely emphasizes categories of major interestPurposely emphasizes categories of major interest
e.g., e.g., Neoplastic ProcessNeoplastic Process No attempt to anything JEPDNo attempt to anything JEPD
No explicit classificatory principles or propertiesNo explicit classificatory principles or properties Textual (not formal) definitionsTextual (not formal) definitions Introduction points for semantic relationshipsIntroduction points for semantic relationships
5 Lister Hill National Center for Biomedical CommunicationsLister Hill National Center for Biomedical Communications
Semantic relationsSemantic relations
Single-inheritance hierarchySingle-inheritance hierarchy Class-class relationsClass-class relations Simply mirrored by inversesSimply mirrored by inverses Weakest reading possible: some-someWeakest reading possible: some-some
Sufficient for some applications (e.g., semantic Sufficient for some applications (e.g., semantic interpretation, reporting and visualization of clinical interpretation, reporting and visualization of clinical information)information)
Too limited for reasoningToo limited for reasoning
6 Lister Hill National Center for Biomedical CommunicationsLister Hill National Center for Biomedical Communications
Semantic groupsSemantic groups
15 collections of semantic types15 collections of semantic types Created for visualization purposesCreated for visualization purposes Purposely non-ontological (not subtrees from the Purposely non-ontological (not subtrees from the
isaisa hierarchy of STs) hierarchy of STs) Based on common properties of (sometimes) Based on common properties of (sometimes)
7 Lister Hill National Center for Biomedical CommunicationsLister Hill National Center for Biomedical Communications
Semantic categorizationSemantic categorization
Generally corresponds to Generally corresponds to isaisa(rarely (rarely is an instance ofis an instance of))
Convenient for extracting a classConvenient for extracting a class Direct access: no traversal necessaryDirect access: no traversal necessary Bypasses hierarchies in vocabularies: not subject to Bypasses hierarchies in vocabularies: not subject to
8 Lister Hill National Center for Biomedical CommunicationsLister Hill National Center for Biomedical Communications
Semantic type assignment (1)Semantic type assignment (1)
Essentially manual (default based on source Essentially manual (default based on source information, reviewed by Metathesaurus editors)information, reviewed by Metathesaurus editors)
Complex and labor intensiveComplex and labor intensive Multiple ST assignment sometimes requiredMultiple ST assignment sometimes required
Structure + role (chemicals)Structure + role (chemicals) Systematic polysemySystematic polysemy
GuidelinesGuidelines Usage notesUsage notes Prior categorization of similar conceptsPrior categorization of similar concepts
9 Lister Hill National Center for Biomedical CommunicationsLister Hill National Center for Biomedical Communications
Semantic type assignment (2)Semantic type assignment (2)
No constraints based on mandatory consistency No constraints based on mandatory consistency between SN and Metathesaurusbetween SN and Metathesaurus(e.g., ST of the child concept must be identical to (e.g., ST of the child concept must be identical to or a descendant of ST of the parent concept)or a descendant of ST of the parent concept)
No constraints based on ontological principles No constraints based on ontological principles (e.g., disjunction between (e.g., disjunction between EntityEntity and and EventEvent))
No constraints based on structural principlesNo constraints based on structural principles(e.g., allowable hybrid types)(e.g., allowable hybrid types)
10 Lister Hill National Center for Biomedical CommunicationsLister Hill National Center for Biomedical Communications
Systematic polysemy (splitting vs. lumping)Systematic polysemy (splitting vs. lumping)
Metathesaurus (RxNorm) distinguishes betweenMetathesaurus (RxNorm) distinguishes between Clinical drug (e.g., Acetaminophen)Clinical drug (e.g., Acetaminophen) Branded drug (e.g., Tylenol)Branded drug (e.g., Tylenol)
But does not systematically distinguish betweenBut does not systematically distinguish between Prostatic adenoma (the tumor responsible for Prostatic adenoma (the tumor responsible for
compressing the urethra)compressing the urethra) Prostatic adenoma (the disease of which urinary Prostatic adenoma (the disease of which urinary
problems are one manifestation)problems are one manifestation)
both contain acetaminophenas their active ingredient
11 Lister Hill National Center for Biomedical CommunicationsLister Hill National Center for Biomedical Communications
FindingFinding
Role played by many different typesRole played by many different types Necessarily some-some (rare exceptions)Necessarily some-some (rare exceptions) Reified for convenienceReified for convenience
12 Lister Hill National Center for Biomedical CommunicationsLister Hill National Center for Biomedical Communications
Overall constraints for changesOverall constraints for changes
Finite amount of resourcesFinite amount of resources Driven by usefulnessDriven by usefulness
Suggestions
14 Lister Hill National Center for Biomedical CommunicationsLister Hill National Center for Biomedical Communications
SN SN andand Metathesaurus Metathesaurus
Issues in the SN cannot be dissociated from issues Issues in the SN cannot be dissociated from issues in the Metathesaurusin the Metathesaurus
Inaccurate/inconsistent concept categorizationInaccurate/inconsistent concept categorization May be a bigger issue than issues identified in the SNMay be a bigger issue than issues identified in the SN
Relatively frequentRelatively frequent Impair semantic integration and semantic interpretationImpair semantic integration and semantic interpretation
Will not be solved solely be addressing issues in the SNWill not be solved solely be addressing issues in the SN
15 Lister Hill National Center for Biomedical CommunicationsLister Hill National Center for Biomedical Communications
SN SN vs.vs. Biomedical ontology Biomedical ontology
Having a good (high-level) ontology of Having a good (high-level) ontology of biomedicine is certainly desirable…biomedicine is certainly desirable…
But it will be of little use if it is not linked to But it will be of little use if it is not linked to Metathesaurus conceptsMetathesaurus concepts
Some ontological features (e.g., some-all) require Some ontological features (e.g., some-all) require a much finer granularity than that of the current a much finer granularity than that of the current semantic typessemantic types
16 Lister Hill National Center for Biomedical CommunicationsLister Hill National Center for Biomedical Communications
Editing vs. AuditingEditing vs. Auditing
Auditing must be pursued, but…Auditing must be pursued, but… Better editing Better editing environmentsenvironments are needed are needed
Law: explicit classificatory principles and propertiesLaw: explicit classificatory principles and properties Order:Order:
Enforce SN/Meta consistencyEnforce SN/Meta consistency(use SN relations as a reference for Meta relations)(use SN relations as a reference for Meta relations)
Restrict allowable combinations of STsRestrict allowable combinations of STs
Quality assurance starts at the time of editingQuality assurance starts at the time of editing
17 Lister Hill National Center for Biomedical CommunicationsLister Hill National Center for Biomedical Communications
Source transparency vs. Anarchy (1)Source transparency vs. Anarchy (1)
All relations asserted by sources are All relations asserted by sources are recordedrecorded……(source transparency)(source transparency)
But need not be necessarily But need not be necessarily trustedtrusted
Similar to how synonymy is treatedSimilar to how synonymy is treated Metathesaurus synonymy does not always follow Metathesaurus synonymy does not always follow
source synonymysource synonymy
18 Lister Hill National Center for Biomedical CommunicationsLister Hill National Center for Biomedical Communications
Source transparency vs. Anarchy (2)Source transparency vs. Anarchy (2)
Similar to how names lacking face validity are Similar to how names lacking face validity are treatedtreated Fully specified Metathesaurus names are createdFully specified Metathesaurus names are created Invalid names are made suppressibleInvalid names are made suppressible
Similarly for relationsSimilarly for relations Metathesaurus hierarchical relations should ignore Metathesaurus hierarchical relations should ignore
some obviously non-hierarchical relations used to some obviously non-hierarchical relations used to create hierarchies in source vocabulariescreate hierarchies in source vocabularies
Suppressibility or Content View Flag (CVF)Suppressibility or Content View Flag (CVF)
Agenda
20 Lister Hill National Center for Biomedical CommunicationsLister Hill National Center for Biomedical Communications
Semantic typesSemantic types
Rename some types (face validity)Rename some types (face validity) Extract explicit classificatory principlesExtract explicit classificatory principles Rearrange hierarchy as needed (e.g., Rearrange hierarchy as needed (e.g., AlgaAlga)) Revisit rolesRevisit roles
Place under sortals when unique (e.g., Place under sortals when unique (e.g., EnzymeEnzyme)) Create allowable hybrids (e.g., Create allowable hybrids (e.g., Steroid hormoneSteroid hormone))
21 Lister Hill National Center for Biomedical CommunicationsLister Hill National Center for Biomedical Communications
Semantic relationsSemantic relations
Align with Metathesaurus relationsAlign with Metathesaurus relations(e.g., (e.g., caused_bycaused_by / / due_todue_to))
Multiple inheritance (?)Multiple inheritance (?) Two levelsTwo levels
Coarse class-class, some-some, with mirrored inversesCoarse class-class, some-some, with mirrored inversesto to labellabel the relation (and support semantic the relation (and support semantic interpretation)interpretation)
Finer non-symmetric class-class, some-all (?)Finer non-symmetric class-class, some-all (?)to support reasoningto support reasoning
22 Lister Hill National Center for Biomedical CommunicationsLister Hill National Center for Biomedical Communications
ST assignmentST assignment
Facilitated by improved editing environmentFacilitated by improved editing environment Driven by explicit classificatory principles and Driven by explicit classificatory principles and
propertiesproperties Simplified by allowable hybridsSimplified by allowable hybrids Constrained by coherence with SN relations Constrained by coherence with SN relations
(requires aligned relations and labeled (requires aligned relations and labeled Metathesaurus relations)Metathesaurus relations)