ontology training examples. what does linked mean? strategy serves retrieval, but not reasoning
TRANSCRIPT
Ontology Training
Examples
What does ‘linked’ mean?’Strategy serves retrieval, but not reasoning•
• poisoning of wells
• no global governance• poor treatment of time
• data and objects confused• uncontrolled proliferation of links
3
What you get with ‘mappings’
All in Human Phenotype Ontology (= all phenotypes: excess hair loss, splayed feet ...)
mapped to
• all organisms in NCBI organism classification
• allose in ChEBI chemistry ontology
• Acute Lymphoblastic Leukemia in National Cancer Institute Thesaurus
4
What you get with ‘mappings’
all phenotypes (excess hair loss, splayed feet ...)
all organisms
allose (a form of sugar)
Acute Lymphoblastic Leukemia
5
Mappings are hard
They are fragile, and expensive to maintainThe goal should be to minimize the need for
mappings
6
Pistoia AllianceOpen standards for data and technology
interfaces in the life science research industry
consortium of major pharmaceutical companies working to address the data silo problems created by multiplicity of proprietary terminologies
declare terminology ‘pre-competitive’
require shared use of OBO Foundry ontologies in presentation of information
http://pistoiaalliance.org/
7
An Example
UCore SL Training SessionANSER Conference Center 2900 Quincy Street, Arlington, Virginia Wednesday, March 17, 2010
Barry Smith and Lowell Vizenor
starts with Continental Breakfast 8am
0900 Administrative Comments (Jim Schoening) 0910 Opening Remarks (Cliff Daus, Clay Robinson DoD CIO) 0930 UCore, the Net-Centric Data Strategy and the Ontological
Perspective (BS) Ontology success stories, and some reasons for failure The promise of the Universal Core Realizing the Net-Centric Data Strategy UCore SL history / team / acknowledgements UCore SL benefits
1030 UCore Semantic Layer: A Logically Enhanced (OWL) Version of the UCore Taxonomy (LV)
Overview of UCore 2.0 TaxonomyOverview of UCore SLUCore SL TaxonomyUCore SL RelationsEquivalence RelationsDisjointness AxiomsRestriction Classes
1130 Effecting Successful Data Coordination (BS) The human factors: traffic rules for ontologistsTop Down / Bottom Up (TDBU) methologyDealing with vocabulary conflicts across communitiesRegistration of metadataTraffic rules for definitionsTraffic rules for relations 1220 Lunch
1330 Applications of UCore SL (BS) Using semantics for quality assurance of UCorePreamble on BFO: RoleThe change management processCreation of coherent extensions of UCoreUCore SL and external resourcesNIEM, C2 Core 1430 Developing Ontologies with UCore SL (LV) How to extend UCore SL How to validate extensions of UCore SLHow to represent temporal qualification of relations 1530 A Strategy for the Future (BS) 1630 Fin
http://ontologist.com
Lecture 7. Towards a Standard Upper Level Ontology
Video • SlidesScientific ontologies have special featuresBuilding scientific ontologies which work together demands a common set of ontological relationsBasic Formal Ontology: benefits of coordinationUsers of BFOContinuants, occurrents, realizablesSpecific dependence, generic dependence, information artifactsDispositions, roles, functionsDiseases and disorders: the Ontology of General Medical Science
HighFleet Training Sample
Our approach is to introduce the ideas and syntax for
• Classes• Relations• Functions• Rules and Integrity Constraints
with hands-on exercises performed by trainees as a single group between sections
© 2010 HighFleet Inc.
© 2010 HighFleet Inc.
Taxonomy - how it worksMammal
Cat
Classes are the buildingblocks of taxonomies
The subsumption relationorganizes classes bygeneralization
When we say that Mammalsubsumes Cat, we mean: a) All Cats are Mammals b) Some Mammals are not Cats
In the ULO, subsumption iscalled ‘sup’, for super-property
sup
Instances of classes are(also) instances of everysuper-class
Reflexive
Symmetric
Transitive
Irreflexive
Anti-symmetric
Asymmetric
If R(x,y) then not R(y,x)
If Steve is better than Bill, than Bill can’t be better than Steve
betterThan(x,y)
If Steve is located in Kansas, then Kansas is not located in Steve
locatedIn(x,y)
tallerThan(x,y)
properPartOf(x,y)
Kinds of Binary Relations
© 2010 HighFleet Inc.
• (and (Dog ?x) (Cat ?y) (chases ?x ?y) (Pet ?y))• (or (nextTo ?x John) (runningFrom ?x Pete))• (=> (Person ?x) (Happy ?x))• (not (between ?x Mary ?y))• (exists (?x) (knows John ?x))• (forall (?x) (=> (knows John ?x) (Horse ?x)))
Rules & Constraints
© 2010 HighFleet Inc.
Training 3: The Pet Store Your client would like you to create an ontology for the entities described below. The instructor will provide any Subject Matter Expert information you request. Create a "Training 3 -- Solution.kfl" file and make an ontology to meet the client's needs. Use only Properties - no Relations or Functions should be used. Use the Training context ------------------------------ CLASSES Dogs, Cats, Birds, Snakes, Fish, Frogs, Lizards, Mice, Mammals, Reptiles, Amphibians, Birds, Fish, Species, Pets, Food, Guard Dogs, People, Customers, Employees, Banned Persons, VIP's, Managers, Suppliers, Mannequins, Plastic Dogs, Plastic Birds, Swedish Fish Candy© 2010 HighFleet Inc.
Training 6: Real Estate Your client would like you to create an ontology for the entities described below. The instructor will provide any Subject Matter Expert information you request. Create a "Training 6 -- Solution.kfl" file and make an ontology to meet the client's needs. Properties, Relations, and Functions may be used. Use the Training context ------------------------------ CLASSES Parcels of Land, Buildings, Houses, Factories, Shops Owned, Leased, and Rented Real Estate (if leased, for how many months?) RELATIONS "owner of (real estate)", "renter of (real estate) ", "groundskeeper of (parcel of land)" “(?) locatedOn (?)", “(?) has price (?)" “(land parcel) has zoning (Residential, Commercial, Industrial)“ “(real estate) leased for (N-many) months”© 2010 HighFleet Inc.
Training 9: Car Manufacturing Your client would like you to create an ontology for the entities described below. The instructor will provide any Subject Matter Expert information you request. Create a "Training 9 -- Solution.kfl" file and make an ontology to meet the client's needs. Properties, Relations, and Functions, Rules, and Integrity Constraints may be used. Use the Training context ------------------------------CLASSES: Manufacturing Companies, Cars, Engines, Tires, Seats RELATIONS"mass of", "max pressure of (tires)", "date of manufacture", "supplied by","volume of (engines)", "color of (seats, cars [assume monochrome]) RULES AND CONSTRAINTSA car should have an engine. Cars that weigh 2500 kg must have an engine with a volume of 80 cubic cm. The tires of blue cars shouldn't be less than 30 psi All Boeing cars are blue with red seats. Every blue car that has all yellow seats was made in 1977.© 2010 HighFleet Inc.