Correcting Misconceptions & misinformation for Users and Developers of Ontologies
- NOTE: some, if not all, of the following points can apply to other systems such as taxonomies, controlled vocabularies, etc.
- Purchase the full presentation here or at this page.
- If you find value in my work, please donate here.
1. The Myth of Mappings. When mappings don't map. Prevent semantic harm in your semantic harmonization activity.
MYTH / FALSITY ❌: mappings/harmonization/alignment activities between terms, vocabularies, ontologies, etc., are always correct, complete or true mappings. A given mapping/harmonizaiton activity will be accurate and guarantee correct mappings between your and other terms, vocabularies, etc.
TRUTH(S) ✅: **There are degrees of mappings, degrees of semantic harmoniztion. Mappings may not be correct or complete. There are a number of considersations and tasks to get accurate mappings and harmonization, but not all activities perform them. **
CAUTION: One vocabulary or ontology might import/adopt/assume/reuse another at some broader degree of abstraction. mapping/harmonization activities may not inform you of this when engaging in mapping/harmonization with your vocabulary. Depending on the activity, this may commit you to things you did not know, agree to, or desire (e.g., definitions, broader categorizations, logical formalizations, philosophical worldviews, etc.). Note that an imported ontology in one that you are mapping to adds more complexity and semantics to the harmonization process. That added content may not be what you desire or agree with. Some ontologies do not disclose this aspect, which will be to the advantage of the imported ontology but to the disadvantage of other. So expect/demand transparency and examine those imported systems. So be aware of any imported, adopted, or reused vocabularies/ontologies that one or both of the mapped/harmonized vocabs/ontologies have or make.
- A mapping or harmonization between two (or more) terms, vocabularies, or ontologies is only a complete and true mapping if all degrees or levels of abstraction, nesting, categorization, subsumption and commitments are understood and examined for possible similariries between the two. (E.g. a parent category in one ontology may not be intended or included in another. the former thereby has more semantic baggage that may not be desired, or that may misdefine the latter).
- Alleged mappings may be false, incorrect, or partial if the intended meaning of the mapped entities is not examined and compared.
- A string match is not enough: a mapping that only looks at the symbol/character string or the natural-language phrasing of the term is a shallow mapping and incomplete. (The Shallowness of Harmonization)
- It is an incomplete mapping if the terms are defined using other (perhaps more abstract) terms or if they have commitments or constraints (e.g., via subsuming or parent category terms). So these aspects must be understood when doing mappings.
- Degrees of mapping/harmonization: there can be exact mapping, partial mapping, no mapping, etc.
- Analogy of the Shallowness of Mappings/Harmonization:
- dipping your toe in a lake is like declaring two leaf terms are semantically equivalent (have the same meaning).
- putting your foot in the like is like declaring the two leaf terms with their immediate parent (more abstract/superordinate/subsuming) terms are semantically equivalent.
- ...and so on depending on how deep or abstract the nesting goes.
- Analogy of the Shallowness of Mappings/Harmonization:
- Compare the intended meaning and formal logical semantics of the two target terms/concepts that potentially map to one another. All commitments, at each degree of abstraction, of each of the two target concepts, terms, symbol-structures or ontologies to be mapped (or semantically harmonizaed) need to examined and compared. E.g., you need to understand the intended meaning and formal logical semantics of the ground (or most specific leaf term) is, and whether or to what extent it's the same as the other term it potentially maps to.
NOTE: 'mapping' and 'semantic harmonization' here means the activity of identifying potential (dis)similarities such as equivalences in intended meaning between two symbol structures (e.g., a term in a vocabulary or ontology, a class with a term label, etc.), and computably asserting them.
MYTH / FALSITY❌: you should use an existing upper- or top-level ontology (also called foundational, core, or generic ontology).
TRUTH ✅: You do not need an upper- or top-level ontology.
- The same results or benefits purported by supporters of uppper ontolgoies can be achieved through other means. And the myth is often used as a strategy for self-gain by those upper groups to acquire more users and thereby more $ potntial.
- You can create your own upper or top-level ontology. You can create it collaboratiely or otherwise.
- NOTE: I encourage this in order to foster creativity, diversity and innovation in the communtiy. It will also help ensure ethical practice by helping prevent the monopolization and unfair advantage of any single existing top-level ontology.
- NOTE: this is not intended to ignore the years of collaborative effort that have gone into many of the existing top-level ontologies. Yes, such organized and team effort is often more efficient, but this myth and the correspding fact is simply to state the facts.
- Re-use is often used as an unethical strategy to justify using someone's ontology product. In reality, re-use depends on the situation: the project and it's specifics, whether you agree on the worldview and semantics of the upper ontology, whether the ontology fits your data and domain expertise, etc.
Myth / Falsity❌: you should re-use existing ontologies (any type).
TRUTH✅: You do not need to (re)use an existing ontology. You may reuse. You can instead create your own and be innovative!
- It depends on the situation. Re-use may be helpful sometimes, but it is not necessary.
- Some who argue for reusing it are assuming that the intended meaning, model and conceptualizations between the potentially (re)used ontology and the user are the same. But we cannot assume this.
- A given ontology (or knowledge model or semantic model or conceptual model) may be too complex.
- If you consider re-using other ontologies, then excersie your right to be infored about the intended meaning, formal semantics, assumptions, and commitments the candidate ontology and its developers. Exercise due diligence so that you do not commit yourself to a model, or a worldview that you don't agree with or that may exclude the entities, content, data or knowledge of your expertise, beliefs, and discipline. (e.g., some may exclude or mischaracterize things like Mind)
With regard to #2, this is why it's important to understand the assumptions and meaning of the most abstract ontology models: because some of them assume philosophical theories that may exclude or mischaracterize things you may want to inlude in your model. For example, some ontologies will deny the existence, or miscategorize, things like mind, truth, God, etc. Some may assume a philosophical theory about activities or processes (very broadly concieved) that you may not agree with, or that conflicts with what your data or domain knowledge may suggest. Some may assume a particular theory of causality (cause-effect) that you do not subscribe to. And so on. This is one reason to make your own ontology and, if desired, your own generic or top-level ontology.
Myth / Falsity❌: Do not re-invent the wheel (reuse existing ontologies).
TRUTH✅: The analogy is too simplistic for ontology engineering. You do not need to, but it may be helpful sometimes. You can create and be innovative!
-
There are ways to achieve interoperability besides re-use.
-
That myth is expressed in various disciplines, often to the advantage of a few who wish to gain users or custoers of their product. It is often used to justify--incorrectly so--the use of a single product or resource, to the disadvantage of other (perhaps competing products). It's abused as a unethical strategy to get one's ontology adopted or used.
-
Tool Re-use vs. Semantic Re-use.
- Re-use is more easily done with tools that are agnostic as to semantics, knowledge, etc.
- BUT, knowledge organization systems and semantic resources such as ontologies are not agnostic: they are inherently about meaning/semantics/knowlege. Some of it may be about profound, sensitive or unanswered concepts of humanity. And therefore there is an ethical aspect and responsibiity to it. Re-use depends on whether that meaning matches yours (or your data), or whether it is correct/accurate/acceptable/etc. We have an moral responsibility to not simply re-use without knowing the intended meaning, assumped worldviews, and commitments made by semantic models like ontologies, metadata sets, vocabularies, etc. You have a right to be aware and be informed, regardless of the degree of abstraction.
Similar to Myth 3 and Myth 4, we have the following myth that is often accompanied by 3 and 4.
Myth / Falsity❌: creating ones own ontology is an obstactle to interoperability. ((m) for short.) TRUTH✅:
- (1) the claim, (m), itself is problematic, requiring far more detail to clarify.
- Problem of interoperability: What sort of interoperability? What is meant by interoperability? For a given sort of interoperability, can it in fact be achieved by the myth or its contrary?
- Ethically Problematic: It is not proper for anyone or any organization to imply or dictate to others that they cannot or should not create their own ontology.
- (2) Creating ones own ontology is NOT an obstacle to interoperability of the appropriate sort.
- It is tantamount to, if not coming from a motivation of, wanting others to use one or more existing ontologies which may or may not be owned by the party who make such statements, theeby being biased, if not ethically questionable. Doing so may be a tactic to get users of a particular existing ontology.
If you follow that line of thinking to its full extent, you will end up in a state of monopolization, reflective of an intentional or unintentional exlcusionaryness (if not elitism) guised by what seems to be noble goals (e.g., interoperability) but whose basic concepts (e.g., that interoperability) are in fact unclear and questionable. From wanting others to use one or more existing ontologies, the line of thinking will involve questioning the existing of even two ontologies of the same topic, resulting in the elimination of one. It will lead you to reducing that number of existing ontologies until there is just one. But then who owns, develops, manages that single, that one ontology? And is such a universal ontology possible, if desired and beneficial at all? Will all people and organizations be able to modify it? Who is the authority(s) for modifications? What if people or orgs. disagree with the content (e.g., the terms, definitions, etc.)?
© 2017-2023, Robert John Rovetto. All right reserved. Not authorized for commercial use unless explicitly negotiated with the author. Citation/attribution required. If you use any content from these Github pages, then you must cite them. No claims to completeness. No liability. No warranty.