Gary,

Answers to your question have been developed many times, often in an ad hoc way.  Short answer:  no single precise detailed definition could be adequate for the kinds of examples you cited below. 

GBC:   How do we avoid infinite regress in an attempt to "standardize" a vocabulary in a complex field?

My recommendation is the solution that Doug Lenat adopted for Cyc, which is similar to the methods we adopted for our VivoMind and Permion companies:

1. Start with a minimally specified top-level ontology.

2. Extend it with broad definitions at the level of WordNet and other lexical resources.

3. For various applications, define detailed ontology modules for the special cases that require high precision.

4. Provide methods for communicating among independently developed modules.  It's possible that some details in some modules cannot be translated with full precision to and from other modules.  The existence of items in those modules may be made known to other modules, but some details might not be exportable outside of the module in which they were defined.

At the Ontology Summit session on Wednesday, Arun and I will show how these methods are used to solve complex problems and to develop long-term solutions.  For an overview of how ontologies were developed and extended for VivoMind applications (from 2000 to 2010), see https://jfsowa.com/talks/cogmem.pdf .

We'll add more info about how Permion Inc.has developed a hybrid system that combines the best of the VivoMind symbolic approach with LLM-based methods for translating languages of any kind -- natural, logical, diagrammatic, multidimensional, and perceptual (as mapped to and from sensory input and physical manifestations).

As an example, one VivoMind customer required the ability to analyze and translate Chinese documents.  An unemployed Chinese linguist (who was raising her children) was very happy to get a job in which she could work at home to develop a Chinese grammar for the VivoMind system.  It worked so well that VivoMind was better able to detect and relate Chinese proper names than the software developed by the Chinese themselves.  (That was in 2010.)

There is no way that a single formal ontology with rigidly defined terms could relate both Chinese and English.  But it is possible to develop an ontology specialized for a particular document (or a limited set of documents) that relates the English terms to and from the Chinese terms in those documents and their English translations.

That was done to the satisfaction of the customer that paid for the development.  Arun and I will discuss these topics on Wednesday.  Ken will announce the talk and the ZOOM address tomorrow.

John