Actions

ConferenceCall 2026 06 10: Difference between revisions

Ontolog Forum

No edit summary
No edit summary
 
(One intermediate revision by the same user not shown)
Line 20: Line 20:
= [[OntologySummit2026|Ontology Summit 2026]] {{#show:{{PAGENAME}}|?session}} =
= [[OntologySummit2026|Ontology Summit 2026]] {{#show:{{PAGENAME}}|?session}} =
* '''Bill Mandrick''' ''Ontology Engineering 101 - From Expert Knowledge to Ontological Models''
* '''Bill Mandrick''' ''Ontology Engineering 101 - From Expert Knowledge to Ontological Models''
** [https://ontologforum.s3.us-east-1.amazonaws.com/OntologySummit2026/Education/Ontology-Engineering-101--BillMandrick_20260610.mp4 Video Recording]
** [https://youtu.be/MbgJqH3-IMQ YouTube Video]
** Abstract: Ontology engineering is a technical field, but at its core it begins with a familiar philosophical task of making distinctions clear. Organizations routinely depend on data whose meaning is only partly understood. Ontology engineering provides a disciplined method for moving from informal expert language to reusable, inspectable, machine-checkable representations. In this talk, Bill Mandrick will present an overview of the weekly Ontology 101 series, 6-week cycles focused on developing foundational ontology skills with hands-on practice. The first two weeks involve participants learning how to work with subject matter experts without trying to turn them into ontologists, the goal being to elicit and refine competency questions: clear, testable questions that the ontology should help answer. In the second two weeks, those questions are translated into visual design patterns that expose the relevant entities, relations, roles, processes, and constraints. In the last pair of weeks, the patterns are implemented in OWL using tools such as Protégé, tested with reasoners, and evaluated against the original competency questions. Throughout the aim is not to master ontology engineering in a single session, but to understand the basic rhythm of the work, engage experts, clarify meaning, model the structure, encode the result, test it, and revise.
** Abstract: Ontology engineering is a technical field, but at its core it begins with a familiar philosophical task of making distinctions clear. Organizations routinely depend on data whose meaning is only partly understood. Ontology engineering provides a disciplined method for moving from informal expert language to reusable, inspectable, machine-checkable representations. In this talk, Bill Mandrick will present an overview of the weekly Ontology 101 series, 6-week cycles focused on developing foundational ontology skills with hands-on practice. The first two weeks involve participants learning how to work with subject matter experts without trying to turn them into ontologists, the goal being to elicit and refine competency questions: clear, testable questions that the ontology should help answer. In the second two weeks, those questions are translated into visual design patterns that expose the relevant entities, relations, roles, processes, and constraints. In the last pair of weeks, the patterns are implemented in OWL using tools such as Protégé, tested with reasoners, and evaluated against the original competency questions. Throughout the aim is not to master ontology engineering in a single session, but to understand the basic rhythm of the work, engage experts, clarify meaning, model the structure, encode the result, test it, and revise.
** Bio: Bill Mandrick, Ph.D. is a senior ontologist at CUBRC and retired U.S. Army Colonel whose work has focused on ontology development, OWL/RDF representation, Basic Formal Ontology compliance, and military/intelligence applications of ontology. Dr. Mandrick is a long-time contributor to early military ontology work and to NCOR/CUBRC best-practices work in ontology development. He also co-authored work with Barry Smith on the philosophical foundations of intelligence collection and analysis, including the role of BFO and the Common Core Ontologies in semantic interoperability for intelligence systems. Dr. Mandrick is the chair of the rather successful "Ontology 101” weekly working group, sponsored by NCOR.  
** Bio: Bill Mandrick, Ph.D. is a senior ontologist at CUBRC and retired U.S. Army Colonel whose work has focused on ontology development, OWL/RDF representation, Basic Formal Ontology compliance, and military/intelligence applications of ontology. Dr. Mandrick is a long-time contributor to early military ontology work and to NCOR/CUBRC best-practices work in ontology development. He also co-authored work with Barry Smith on the philosophical foundations of intelligence collection and analysis, including the role of BFO and the Common Core Ontologies in semantic interoperability for intelligence systems. Dr. Mandrick is the chair of the rather successful "Ontology 101” weekly working group, sponsored by NCOR.  
Line 31: Line 33:


== Discussion ==
== Discussion ==
10:15:52 Mike Bennett: Why are regulated products a subclass? Being regulated is not an inherent property of a product.
10:23:13 Victor.Bagwell: I'm a data guy, so to speak.  So I view this as analogous to relational database normalization to 3rd normal down to dimension tables where the normalization passes are repeated until each hierarchy goes down to mutual exclusive components.  Albeit, going a bit beyond how to store/access the data to ask what exists in reality and "how" they are related.  Essentially a semantic normalization.
10:24:00 Paul Tyson: Session idea: unbiased head-to-head comparison of CL and OWL with working examples and applications. Does anyone know of such a resource already available?
10:26:59 Bill Mandrick: william.mandrick@cubrc.org
10:27:19 Gary Berg-Cross: I find that you can use, as a starting point an AGI agent to find definitions for things not in an Ontology.  For example five folate-related chemicals. Here are their 2 definitions:
# Pteroylmonoglutamic Acid The synthetic form of vitamin B9 (folic acid), representing the first molecule in an enzymatic process that results in the bioactive form of folate. It has high bioavailability and is the only folate form authorized in fortified foods and drugs. Rootine
# 5-Methyltetrahydrofolate (5-MTHF) The main food folate and principal form of vitamin B9 found in plasma. It is the end product of folate metabolism and is involved in the remethylation of homocysteine to methionine, a critical step in methionine and DNA synthesis. ScienceDirectDoveMed
# Tetrahydrofolate (THF) A derivative of vitamin B9 and a coenzyme for metabolic reactions involving amino acid and nucleic acid formations. It participates in important single-carbon transfer reactions — often referred to as one-carbon metabolism — and in synthesizing several amino acids such as serine and methionine, purines, and thymine. Chemically, it consists of three structural components: para-aminobenzoic acid (PABA), a bicyclic pteridine ring, and glutamic acid. NCBI
# L-Methylfolate (5-MTHF) The bioactive, naturally occurring form of folate. It is the metabolic end point of the folate cycle, produced when the MTHFR enzyme converts 5,10-methylenetetrahydrofolate into 5-methyltetrahydrofolate. It is the only form of folate that can cross the blood-brain barrier. MDPI
10:29:17 Gary Berg-Cross: The Defs can then be axiomatized
* Object Properties — 9 semantic relations including has_metabolic_precursor, donates_group, converted_by_enzyme, is_active_form_of, and crosses_barrier, all with inverse declarations where applicable.
* Data Properties — 9 annotation/data properties covering molecular formula, molar mass, CAS number, ChEBI/PubChem IDs, IUPAC name, and boolean flags like is_synthetic and crosses_blood_brain_barrier.
* Class Hierarchy: ChemicalEntity → VitaminCompound → FolateCompound → CoenzymeFolate / SyntheticFolate
** Each of the five compounds as a named class with full OWL restriction axioms
* Key Axioms per compound:
** PteroylmonoglutamicAcid — synthetic, fully oxidised, is_metabolic_precursor_of THF, converted by DHFR
** Tetrahydrofolate — active form, OneCarbon_Carrier, precursor to both 5-MTHF and 10-CHO-THF
** FiveMethyltetrahydrofolate — MethylGroupDonor, crosses BBB, participates in homocysteine remethylation, converted by MTHFR
** LMethylfolate — declared owl:equivalentClass to 5-MTHF (same entity, different clinical name)
** TenFormyltetrahydrofolate — FormylGroupDonor at N10, drives purine synthesis
*General Class Axioms (GCIs) — closure axioms enforcing that only 5-MTHF crosses the BBB within this compound set, and that N10-formyl donation entails membership in TenFormyltetrahydrofolate.
10:30:32 Gary Berg-Cross: Here's a start on an ontology for your diagram. Here's the full ontology, directly loadable in Protégé or any OWL 2 DL reasoner (HermiT, Pellet, ELK). Here's a breakdown of what was axiomatized from the diagram:
* Upper Ontology (BFO-aligned, left spine of the graph) The full BFO chain is reproduced: Entity → Continuant → IndependentContinuant → MaterialEntity → Object → PortionOfProcessedMaterial, with BFO IRIs annotated on each class. The right spine follows Continuant → GenericallDependentContinuant → InformationContentEntity → DirectiveInformationContentEntity.
* Object Properties (11 named relations) All relations visible in the diagram are formally declared — has_ingredient, prescribed_by / prescribes, complies_with, governed_by / governs, specifies / specified_by, has_part / part_of (BFO-aligned), and concretizes / is_concretized_by — with domain/range documentation and inverses where applicable.
* Domain Class Hierarchy (two branches)
** Material branch: PortionOfProcessedMaterial → RegulatedProduct → SupplementProduct → VitaminSupplementProduct (with VitaminIngredient as a sibling)
** Directive branch: PerformanceSpecification → IntendedUseStatement → SupplementUseStatement; QualitySpecification → IngredientSpecification; ProcessRegulation → RegulatoryFramework → SupplementRegulatoryFramework → UruguaySupplementRegulatoryFramework
* Instance Layer — all four pink-bordered individuals from the diagram are instantiated with their exact named relations: VitaminSupplementProduct001 has_ingredient VitaminIngredient001, IngredientSpecification001 prescribes SupplementUseStatement001, and the Uruguay framework instance governs the product and prescribes the spec.
* Disjointness, GCIs,....
* Property Chains enforce closure: e.g., only 5-MTHF crosses the BBB (from the folate ontology pattern), anything governed by a Uruguay framework is a SupplementProduct, and a has_ingredient o specified_by chain propagates prescribed_by transitively.
* The diagram faithfully reproduces the ontograph structure.  to read it:
** Color encoding gray boxes are BFO upper ontology (entity, continuant, independent/generically dependent continuant); blue is the material branch (material entity → object); teal is the information branch (ICE → Directive ICE and its three direct subclasses); purple is the domain layer (all the supplement-specific classes); amber with a heavier border marks the 5 named instances at the bottom.
** Left spine follows the BFO material chain down to Portion of Processed Material, which splits into Regulated Product (leading to Supplement Product → Vitamin Supplement Product) & Vitamin Ingredient side by side.
* Right spine descends from Generically Dependent Continuant → ICE → Directive ICE, which fans into three parallel columns: Performance Specification → Intended Use Statement → Supplement Use Statement; Quality Specification → Ingredient Specification; and Process Regulation → Regulatory Framework → Supplement Regulatory Framework → Uruguay Supplement Regulatory Framework.
* Instance layer (below the dashed separator) shows all five individuals with their named object-property relations rendered as colored dashed arrows: has ingredient, specified by, prescribed by, governed by, complies with, prescribes, and documents. Every node is clickable for deeper explanation.
10:28:12 Michael DeBellis: Since you're using Web Protege you couldn't define any axioms on classes. Do you have a process where you do further elaboration on the model and add axioms?
10:37:01 Paul A. Pope: Could non-structural characterizations, like "regulated", be placed in the Annotations attached to a class?  Perhaps in a Description annotation?  "Description: Regulated" or "Regulated: Yes"  (BTW, I don't have mic capability)
10:40:45 Michael DeBellis: When modeling a domain, IMO you seldom use the natural language definition. That's far too expansive. In a general NLP ontology speed and velocity are synonyms. In a physics ontology velocity is a vector and speed is a scalar. The same is true for classes like product, purchase order, etc.
10:42:19 TS: In natural language processing the corpus of material processed provides the context for interpretation.
10:50:48 Victor.Bagwell: reponse to Michael/TS...just for clarity -- My comments were related to a foundational corpus.  I've been involved in NLP for sometime and understand the specific effects of domains (specific to).  Rather, I was thinking about public common use case at the top (e.g., websters) and what the prob. is of semantics for a specific word.  Then a hierarchy that includes are is split (perhaps by domain), and adjusted for the hierarchical levels by the domain, sub-domain, sub-sub-domain, n as a cross section and then over time and perhaps even by other compnents (e.g., language, education culture, geography, to name a few).
Ultimately -- highly dimensional
I was thinking about how to operational information in real context
10:52:04 Paul A. Pope: (circling back here)  Could non-structural characterizations, like "regulated", be placed in the Annotations attached to a class? Perhaps in a Description annotation? "Description: Regulated" or "Regulated: Yes"
* Thank you for addressing my question.  Great discussion.  Looking forward to subsequent meetings.
11:01:37 Gary Berg-Cross: We have to define what subsequent meetings we will have.
11:11:34 Victor.Bagwell: Thank you!  Always learning from all of you.
* Marcia Zeng: 👍


== Resources ==
== Resources ==
* [https://ontologforum.s3.us-east-1.amazonaws.com/OntologySummit2026/Education/Ontology-Engineering-101--BillMandrick_20260610.mp4 Video Recording]
* [https://youtu.be/MbgJqH3-IMQ YouTube Video]


== Previous Meetings ==
== Previous Meetings ==

Latest revision as of 22:22, 10 June 2026

Session Education
Duration 1 hour
Date/Time 10 June 2026 16:00 GMT
9:00am PDT/12:00pm EDT
5:00pm BST/6:00pm CST
Convener Ken Baclawski

Ontology Summit 2026 Education

  • Bill Mandrick Ontology Engineering 101 - From Expert Knowledge to Ontological Models
    • Video Recording
    • YouTube Video
    • Abstract: Ontology engineering is a technical field, but at its core it begins with a familiar philosophical task of making distinctions clear. Organizations routinely depend on data whose meaning is only partly understood. Ontology engineering provides a disciplined method for moving from informal expert language to reusable, inspectable, machine-checkable representations. In this talk, Bill Mandrick will present an overview of the weekly Ontology 101 series, 6-week cycles focused on developing foundational ontology skills with hands-on practice. The first two weeks involve participants learning how to work with subject matter experts without trying to turn them into ontologists, the goal being to elicit and refine competency questions: clear, testable questions that the ontology should help answer. In the second two weeks, those questions are translated into visual design patterns that expose the relevant entities, relations, roles, processes, and constraints. In the last pair of weeks, the patterns are implemented in OWL using tools such as Protégé, tested with reasoners, and evaluated against the original competency questions. Throughout the aim is not to master ontology engineering in a single session, but to understand the basic rhythm of the work, engage experts, clarify meaning, model the structure, encode the result, test it, and revise.
    • Bio: Bill Mandrick, Ph.D. is a senior ontologist at CUBRC and retired U.S. Army Colonel whose work has focused on ontology development, OWL/RDF representation, Basic Formal Ontology compliance, and military/intelligence applications of ontology. Dr. Mandrick is a long-time contributor to early military ontology work and to NCOR/CUBRC best-practices work in ontology development. He also co-authored work with Barry Smith on the philosophical foundations of intelligence collection and analysis, including the role of BFO and the Common Core Ontologies in semantic interoperability for intelligence systems. Dr. Mandrick is the chair of the rather successful "Ontology 101” weekly working group, sponsored by NCOR.

Conference Call Information

Discussion

10:15:52 Mike Bennett: Why are regulated products a subclass? Being regulated is not an inherent property of a product.

10:23:13 Victor.Bagwell: I'm a data guy, so to speak. So I view this as analogous to relational database normalization to 3rd normal down to dimension tables where the normalization passes are repeated until each hierarchy goes down to mutual exclusive components. Albeit, going a bit beyond how to store/access the data to ask what exists in reality and "how" they are related. Essentially a semantic normalization.

10:24:00 Paul Tyson: Session idea: unbiased head-to-head comparison of CL and OWL with working examples and applications. Does anyone know of such a resource already available?

10:26:59 Bill Mandrick: william.mandrick@cubrc.org

10:27:19 Gary Berg-Cross: I find that you can use, as a starting point an AGI agent to find definitions for things not in an Ontology. For example five folate-related chemicals. Here are their 2 definitions:

  1. Pteroylmonoglutamic Acid The synthetic form of vitamin B9 (folic acid), representing the first molecule in an enzymatic process that results in the bioactive form of folate. It has high bioavailability and is the only folate form authorized in fortified foods and drugs. Rootine
  2. 5-Methyltetrahydrofolate (5-MTHF) The main food folate and principal form of vitamin B9 found in plasma. It is the end product of folate metabolism and is involved in the remethylation of homocysteine to methionine, a critical step in methionine and DNA synthesis. ScienceDirectDoveMed
  3. Tetrahydrofolate (THF) A derivative of vitamin B9 and a coenzyme for metabolic reactions involving amino acid and nucleic acid formations. It participates in important single-carbon transfer reactions — often referred to as one-carbon metabolism — and in synthesizing several amino acids such as serine and methionine, purines, and thymine. Chemically, it consists of three structural components: para-aminobenzoic acid (PABA), a bicyclic pteridine ring, and glutamic acid. NCBI
  4. L-Methylfolate (5-MTHF) The bioactive, naturally occurring form of folate. It is the metabolic end point of the folate cycle, produced when the MTHFR enzyme converts 5,10-methylenetetrahydrofolate into 5-methyltetrahydrofolate. It is the only form of folate that can cross the blood-brain barrier. MDPI

10:29:17 Gary Berg-Cross: The Defs can then be axiomatized

  • Object Properties — 9 semantic relations including has_metabolic_precursor, donates_group, converted_by_enzyme, is_active_form_of, and crosses_barrier, all with inverse declarations where applicable.
  • Data Properties — 9 annotation/data properties covering molecular formula, molar mass, CAS number, ChEBI/PubChem IDs, IUPAC name, and boolean flags like is_synthetic and crosses_blood_brain_barrier.
  • Class Hierarchy: ChemicalEntity → VitaminCompound → FolateCompound → CoenzymeFolate / SyntheticFolate
    • Each of the five compounds as a named class with full OWL restriction axioms
  • Key Axioms per compound:
    • PteroylmonoglutamicAcid — synthetic, fully oxidised, is_metabolic_precursor_of THF, converted by DHFR
    • Tetrahydrofolate — active form, OneCarbon_Carrier, precursor to both 5-MTHF and 10-CHO-THF
    • FiveMethyltetrahydrofolate — MethylGroupDonor, crosses BBB, participates in homocysteine remethylation, converted by MTHFR
    • LMethylfolate — declared owl:equivalentClass to 5-MTHF (same entity, different clinical name)
    • TenFormyltetrahydrofolate — FormylGroupDonor at N10, drives purine synthesis
  • General Class Axioms (GCIs) — closure axioms enforcing that only 5-MTHF crosses the BBB within this compound set, and that N10-formyl donation entails membership in TenFormyltetrahydrofolate.

10:30:32 Gary Berg-Cross: Here's a start on an ontology for your diagram. Here's the full ontology, directly loadable in Protégé or any OWL 2 DL reasoner (HermiT, Pellet, ELK). Here's a breakdown of what was axiomatized from the diagram:

  • Upper Ontology (BFO-aligned, left spine of the graph) The full BFO chain is reproduced: Entity → Continuant → IndependentContinuant → MaterialEntity → Object → PortionOfProcessedMaterial, with BFO IRIs annotated on each class. The right spine follows Continuant → GenericallDependentContinuant → InformationContentEntity → DirectiveInformationContentEntity.
  • Object Properties (11 named relations) All relations visible in the diagram are formally declared — has_ingredient, prescribed_by / prescribes, complies_with, governed_by / governs, specifies / specified_by, has_part / part_of (BFO-aligned), and concretizes / is_concretized_by — with domain/range documentation and inverses where applicable.
  • Domain Class Hierarchy (two branches)
    • Material branch: PortionOfProcessedMaterial → RegulatedProduct → SupplementProduct → VitaminSupplementProduct (with VitaminIngredient as a sibling)
    • Directive branch: PerformanceSpecification → IntendedUseStatement → SupplementUseStatement; QualitySpecification → IngredientSpecification; ProcessRegulation → RegulatoryFramework → SupplementRegulatoryFramework → UruguaySupplementRegulatoryFramework
  • Instance Layer — all four pink-bordered individuals from the diagram are instantiated with their exact named relations: VitaminSupplementProduct001 has_ingredient VitaminIngredient001, IngredientSpecification001 prescribes SupplementUseStatement001, and the Uruguay framework instance governs the product and prescribes the spec.
  • Disjointness, GCIs,....
  • Property Chains enforce closure: e.g., only 5-MTHF crosses the BBB (from the folate ontology pattern), anything governed by a Uruguay framework is a SupplementProduct, and a has_ingredient o specified_by chain propagates prescribed_by transitively.
  • The diagram faithfully reproduces the ontograph structure. to read it:
    • Color encoding gray boxes are BFO upper ontology (entity, continuant, independent/generically dependent continuant); blue is the material branch (material entity → object); teal is the information branch (ICE → Directive ICE and its three direct subclasses); purple is the domain layer (all the supplement-specific classes); amber with a heavier border marks the 5 named instances at the bottom.
    • Left spine follows the BFO material chain down to Portion of Processed Material, which splits into Regulated Product (leading to Supplement Product → Vitamin Supplement Product) & Vitamin Ingredient side by side.
  • Right spine descends from Generically Dependent Continuant → ICE → Directive ICE, which fans into three parallel columns: Performance Specification → Intended Use Statement → Supplement Use Statement; Quality Specification → Ingredient Specification; and Process Regulation → Regulatory Framework → Supplement Regulatory Framework → Uruguay Supplement Regulatory Framework.
  • Instance layer (below the dashed separator) shows all five individuals with their named object-property relations rendered as colored dashed arrows: has ingredient, specified by, prescribed by, governed by, complies with, prescribes, and documents. Every node is clickable for deeper explanation.

10:28:12 Michael DeBellis: Since you're using Web Protege you couldn't define any axioms on classes. Do you have a process where you do further elaboration on the model and add axioms?

10:37:01 Paul A. Pope: Could non-structural characterizations, like "regulated", be placed in the Annotations attached to a class? Perhaps in a Description annotation? "Description: Regulated" or "Regulated: Yes" (BTW, I don't have mic capability)

10:40:45 Michael DeBellis: When modeling a domain, IMO you seldom use the natural language definition. That's far too expansive. In a general NLP ontology speed and velocity are synonyms. In a physics ontology velocity is a vector and speed is a scalar. The same is true for classes like product, purchase order, etc.

10:42:19 TS: In natural language processing the corpus of material processed provides the context for interpretation.

10:50:48 Victor.Bagwell: reponse to Michael/TS...just for clarity -- My comments were related to a foundational corpus. I've been involved in NLP for sometime and understand the specific effects of domains (specific to). Rather, I was thinking about public common use case at the top (e.g., websters) and what the prob. is of semantics for a specific word. Then a hierarchy that includes are is split (perhaps by domain), and adjusted for the hierarchical levels by the domain, sub-domain, sub-sub-domain, n as a cross section and then over time and perhaps even by other compnents (e.g., language, education culture, geography, to name a few).

Ultimately -- highly dimensional

I was thinking about how to operational information in real context

10:52:04 Paul A. Pope: (circling back here) Could non-structural characterizations, like "regulated", be placed in the Annotations attached to a class? Perhaps in a Description annotation? "Description: Regulated" or "Regulated: Yes"

  • Thank you for addressing my question. Great discussion. Looking forward to subsequent meetings.

11:01:37 Gary Berg-Cross: We have to define what subsequent meetings we will have.

11:11:34 Victor.Bagwell: Thank you! Always learning from all of you.

  • Marcia Zeng: 👍

Resources

Previous Meetings

 Session
ConferenceCall 2026 06 03Cognition
ConferenceCall 2026 05 20Education
ConferenceCall 2026 05 13Interoperability
... further results

Next Meetings

 Session
ConferenceCall 2026 06 17Synthesis
ConferenceCall 2026 06 24Symposium
Retrieved from "https://ontologforum.com/index.php?title=ConferenceCall_2026_06_10&oldid=5601"