![Page 1: KMi-SMI collaboration Wenjin Lu and Enrico Motta Knowledge Media Institute Monica Crubézy Stanford Medical Informatics](https://reader036.vdocuments.mx/reader036/viewer/2022062516/56649d815503460f94a65ac8/html5/thumbnails/1.jpg)
KMi-SMI collaboration
Wenjin Lu and Enrico MottaKnowledge Media Institute
Monica CrubézyStanford Medical Informatics
![Page 2: KMi-SMI collaboration Wenjin Lu and Enrico Motta Knowledge Media Institute Monica Crubézy Stanford Medical Informatics](https://reader036.vdocuments.mx/reader036/viewer/2022062516/56649d815503460f94a65ac8/html5/thumbnails/2.jpg)
Goals
• Customizing Protégé editor for use with OCML (Protégé-OCML)– Evaluating Protégé suitability
• Translating classification library into Protégé-OCML– Validating classification library– Evaluating UPML
• Revising Specification for Internet Reasoning Service
![Page 3: KMi-SMI collaboration Wenjin Lu and Enrico Motta Knowledge Media Institute Monica Crubézy Stanford Medical Informatics](https://reader036.vdocuments.mx/reader036/viewer/2022062516/56649d815503460f94a65ac8/html5/thumbnails/3.jpg)
1. The Protégé-UPML editor for OCML
• Modeling of OCML meta ontology in Protégé– i.e., classes, relations, functions, axioms, etc...
• Inclusion of OCML meta ontology in Protégé-UPML editor– i.e., defining the appropriate KA forms for the basic OCML
constructs
• Extension of UPML concepts for OCML, with reference to OCML meta ontology– e.g., formula -> OCML-formula
![Page 4: KMi-SMI collaboration Wenjin Lu and Enrico Motta Knowledge Media Institute Monica Crubézy Stanford Medical Informatics](https://reader036.vdocuments.mx/reader036/viewer/2022062516/56649d815503460f94a65ac8/html5/thumbnails/4.jpg)
OCML meta ontology in ProtégéBy the way, this is the output of a new plugin for Protégé,
that creates a graph out of a KB,with different visualization options
subclass-of linkinstance-of link
![Page 5: KMi-SMI collaboration Wenjin Lu and Enrico Motta Knowledge Media Institute Monica Crubézy Stanford Medical Informatics](https://reader036.vdocuments.mx/reader036/viewer/2022062516/56649d815503460f94a65ac8/html5/thumbnails/5.jpg)
Each OCML construct is an instance of :OCML-CLASS
Template slots define own OCML
fields for all classes
![Page 6: KMi-SMI collaboration Wenjin Lu and Enrico Motta Knowledge Media Institute Monica Crubézy Stanford Medical Informatics](https://reader036.vdocuments.mx/reader036/viewer/2022062516/56649d815503460f94a65ac8/html5/thumbnails/6.jpg)
OCML in Protégé-UPML editor
• Inclusion of OCML meta ontology in UPML domain modelling meta-ontology
• Extension of UPML concepts for OCML (“UPML-OCML concepts”)
– Formula < OCML-Formula • formula -> ocml-kappa-expression, ocml-relation-mapping
– Program < OCML-Program• program -> ocml-lambda-expression, :OCML-PROCEDURE
– Signature Element < OCML-Element • ocml-type -> :OCML-CLASS, :OCML-RELATION, :OCML-FUNCTION
– Signature < OCML-Signature • signature-elements -> (OCML-Element)
![Page 7: KMi-SMI collaboration Wenjin Lu and Enrico Motta Knowledge Media Institute Monica Crubézy Stanford Medical Informatics](https://reader036.vdocuments.mx/reader036/viewer/2022062516/56649d815503460f94a65ac8/html5/thumbnails/7.jpg)
UPML core ontology isextended for OCML
![Page 8: KMi-SMI collaboration Wenjin Lu and Enrico Motta Knowledge Media Institute Monica Crubézy Stanford Medical Informatics](https://reader036.vdocuments.mx/reader036/viewer/2022062516/56649d815503460f94a65ac8/html5/thumbnails/8.jpg)
2. The Classification library in UPML
Task-DomainBridge
PSM-DomainBridge
PSM-TaskBridge
Task
Task Refiner
DomainModel
Domain Refiner
PSM
PSM Refiner
Ontologies
Ontology Refiner
Classification Library
Heuristicclassifier
(Optimal)Classification
Optimal classification@ Heuristic classifier
Apples
Heuristic classifier@ Apples
Optimal classification@ Apples
Abtractor,Refiner
Color “green”,Sugar-level
Observable, Solution
Abstractor = Sugar-abstractorSolution = Apple class
![Page 9: KMi-SMI collaboration Wenjin Lu and Enrico Motta Knowledge Media Institute Monica Crubézy Stanford Medical Informatics](https://reader036.vdocuments.mx/reader036/viewer/2022062516/56649d815503460f94a65ac8/html5/thumbnails/9.jpg)
The Classification library in Protégé-UPML
• Classification ontologies modeled in Protégé-OCML– Classification task, heuristic classification method, apples
domain, apples @ heuristic classification application– Modeled as separate ontologies in Protégé-OCML (ie, a
hierarchy of classes that are instances of OCML primitives)– Included along with UPML-OCML (UPML extended for OCML)
• UPML-OCML concepts instantiated with elements of the included classification ontologies
• UPML concepts instantiated with classification tasks, PSMs, knowledge roles, etc.
![Page 10: KMi-SMI collaboration Wenjin Lu and Enrico Motta Knowledge Media Institute Monica Crubézy Stanford Medical Informatics](https://reader036.vdocuments.mx/reader036/viewer/2022062516/56649d815503460f94a65ac8/html5/thumbnails/10.jpg)
Classification ontologies in Protégé-OCML
OCML-BASE
CLASSIFICATION-TASK
HEURISTIC-CLASSIFICATION-METHOD
APPLE-DOMAIN
APPLE-HEURISTIC-CLASSIFICATION-APPLICATION
included-inincluded-in
included-inincluded-in
included-in
![Page 11: KMi-SMI collaboration Wenjin Lu and Enrico Motta Knowledge Media Institute Monica Crubézy Stanford Medical Informatics](https://reader036.vdocuments.mx/reader036/viewer/2022062516/56649d815503460f94a65ac8/html5/thumbnails/11.jpg)
Each concept is an instance of
:OCML-CLASS
Own slots contain OCMLdefinition of the class
![Page 12: KMi-SMI collaboration Wenjin Lu and Enrico Motta Knowledge Media Institute Monica Crubézy Stanford Medical Informatics](https://reader036.vdocuments.mx/reader036/viewer/2022062516/56649d815503460f94a65ac8/html5/thumbnails/12.jpg)
Classification Library instance
![Page 13: KMi-SMI collaboration Wenjin Lu and Enrico Motta Knowledge Media Institute Monica Crubézy Stanford Medical Informatics](https://reader036.vdocuments.mx/reader036/viewer/2022062516/56649d815503460f94a65ac8/html5/thumbnails/13.jpg)
Classification Task instance
![Page 14: KMi-SMI collaboration Wenjin Lu and Enrico Motta Knowledge Media Institute Monica Crubézy Stanford Medical Informatics](https://reader036.vdocuments.mx/reader036/viewer/2022062516/56649d815503460f94a65ac8/html5/thumbnails/14.jpg)
Task: Classification (ontology signature)
![Page 15: KMi-SMI collaboration Wenjin Lu and Enrico Motta Knowledge Media Institute Monica Crubézy Stanford Medical Informatics](https://reader036.vdocuments.mx/reader036/viewer/2022062516/56649d815503460f94a65ac8/html5/thumbnails/15.jpg)
Task: Classification (competence)
![Page 16: KMi-SMI collaboration Wenjin Lu and Enrico Motta Knowledge Media Institute Monica Crubézy Stanford Medical Informatics](https://reader036.vdocuments.mx/reader036/viewer/2022062516/56649d815503460f94a65ac8/html5/thumbnails/16.jpg)
Heuristic optimal solution classifier PSM
![Page 17: KMi-SMI collaboration Wenjin Lu and Enrico Motta Knowledge Media Institute Monica Crubézy Stanford Medical Informatics](https://reader036.vdocuments.mx/reader036/viewer/2022062516/56649d815503460f94a65ac8/html5/thumbnails/17.jpg)
Heuristic optimal solution classifier PSM (Signature and Competence)
![Page 18: KMi-SMI collaboration Wenjin Lu and Enrico Motta Knowledge Media Institute Monica Crubézy Stanford Medical Informatics](https://reader036.vdocuments.mx/reader036/viewer/2022062516/56649d815503460f94a65ac8/html5/thumbnails/18.jpg)
Heuristic optimal solution classifier PSM (Operational description)
![Page 19: KMi-SMI collaboration Wenjin Lu and Enrico Motta Knowledge Media Institute Monica Crubézy Stanford Medical Informatics](https://reader036.vdocuments.mx/reader036/viewer/2022062516/56649d815503460f94a65ac8/html5/thumbnails/19.jpg)
Classification-task @ apples-domain bridge
4 mapping axioms define the bridge
![Page 20: KMi-SMI collaboration Wenjin Lu and Enrico Motta Knowledge Media Institute Monica Crubézy Stanford Medical Informatics](https://reader036.vdocuments.mx/reader036/viewer/2022062516/56649d815503460f94a65ac8/html5/thumbnails/20.jpg)
Classification-task @ apples-domain bridge (mapping axiom for observable)
![Page 21: KMi-SMI collaboration Wenjin Lu and Enrico Motta Knowledge Media Institute Monica Crubézy Stanford Medical Informatics](https://reader036.vdocuments.mx/reader036/viewer/2022062516/56649d815503460f94a65ac8/html5/thumbnails/21.jpg)
Conclusions of the experiment
• Customization of Protégé-UPML for OCML– 2-step process: OCML in Protégé + extension of UPML for
OCML– Reasonably easy to do!
• Classification library in UPML– Fits nicely, except:
• Knowledge roles (case-indep. data) at the task & complex PSM level
• Task, PSMs, etc. are classes in OCML, not instances as in UPML
• Needs the implementation of a connection to OCML interpreter
![Page 22: KMi-SMI collaboration Wenjin Lu and Enrico Motta Knowledge Media Institute Monica Crubézy Stanford Medical Informatics](https://reader036.vdocuments.mx/reader036/viewer/2022062516/56649d815503460f94a65ac8/html5/thumbnails/22.jpg)
3. Internet Reasoning Service
![Page 23: KMi-SMI collaboration Wenjin Lu and Enrico Motta Knowledge Media Institute Monica Crubézy Stanford Medical Informatics](https://reader036.vdocuments.mx/reader036/viewer/2022062516/56649d815503460f94a65ac8/html5/thumbnails/23.jpg)