Controlled vocabularies definition methodfor bridging formal ontologies

development

Paolo Ciccarese, PhDMass General Hospital / Harvard Medical School

Background

• AlzForum http://www.alzforum.org/• Semantic Web Applications in Neuromedicine

SWAN: http://swan.mindinformatics.org/SWAN Alzheimer: http://hypothesis.alzforum.org/

• Science Collaboration FrameworkSCF: http://www.sciencecollaboration.org/StemBook: http://www.stembook.org/

• HCLSIG Scientific Discourse Task Forcehttp://esw.w3.org/topic/HCLSIG/SWANSIOC

time

SWAN Ontology Ecosystem

Domain Ontologies

Qualifiers

Basic

An example of SWAN content curationSeeding neuritic plaques from

the distance: a possible role for brainstem neurons in the

development of Alzheimer's disease pathology.

Muresan Z, Muresan V

SWAN CuratorGwen Wong, PhD

Journal Article

SWAN Workbench

Triple store

SWAN Browser

CommentsQualifiers

Kim hypothesis

Graph view

An example of Research Statement

Relationships between discourse elements

Life Science Entities - 1Research Statement

refersTo

refersTo

Life Science Entities - 2Research Statement

But…

• Creating ontologies takes time long time• Creating ontologies requires skilled

knowledge engineers and skilleddomain experts

• The time needed for developingontologies is influencing the applicationdevelopment

• Is creating ontologies always theoptimal solution?

SWAN Additional Annotation - 1Research Statement

qualifiedBy

qualifiedBy

Mechanisms Taxonomy

Pathogenic Narrative

Hypothesis

Claim

qualifiedBy

Other Additional Annotation - 2

* Nature Reports - Stem Cells Cheat SheetNature Publishing Group

Need for controlled vocabularies

• Using terms coming from controlledvocabularies is a good compromise inthe case of high quality human curatedscientific content

• Maybe we can use controlledvocabularies as an easy andincremental method to get to formalontologies

Requirements

• Controlled vocabularies• Controlled vocabularies should be easy and

fast to be defined by “not ontologists”• Link the terms to existing (or future)

ontologies and keep track ofprovenance/authoring

• OWL-DL but without impacting thereasoning

• Share the vocabularies in RDF format

Agile definition and mapping

• We don’t want users “not ontologists” todeal with classes, properties andrestrictions

• Taxonomies are often enough forsimple classifications and can bedefined easily (terms + definitions)

• We want trained ontologists to figureout how to map the terms to existingontologies (if we have resources to doso)

Possible solution: SKOSThe Simple Knowledge Organization System provides amodel for expressing the basic structure and content ofconcept schemes such as thesauri, classification schemes,subject heading lists, taxonomies, folksonomies, and othersimilar types of controlled vocabulary.

How to refer to formal ontologies?

? ?

An example

er

skos:Concept qualifiers:Meaning

owl:Thing

qualifiers:exactMeaningqualifiers:broaderMeaningqualifiers:narrowerMeaning

qualifiers:meaningURI

Mapping to formal ontology

HSC Haematopoietic stem cells (blood-forming stem cells that reside in bone

marrow) 2009-01-28T00:00:00+05:00 2009-01-29T10:00:00+05:00 Paolo Ciccarese

Meaning Of A Tag (MOAT) - 1

Meaning Of A Tag (MOAT) - 2

http://moat-project.org/

Alignment with MOATMeaning

moat:Meaningqualifiers:Meaning

moat:hasMeaning

hasMeaning

qualifiers:hasExactMeaning

meaningURI

moat:meaningURIqualifiers:meaningURI

range

domain

ConclusionsEasy to implement (alsoat application level)OWL-DL (with OWL2 -punning)ProvenanceMapping to existingontologiesDoesn’t changereasoning (unless we post-process the annotation)Aligned with MOAT

Introduces a level ofcomplexity (but wecan always post-process theannotation)