Defining criteria for accepting precomposed terms to avoid combinatorial bombs!

[ddooley]

Need to have design principle on when to have precomposed terms vs. ontology design pattern for expressing a patterned entity. Potential factors:

• Market demand:
  • Commercial / research database / marketplace usage
• Use-case questions it can play a role in answering
• How complex is knowledge graph structure of non-precomposed term?
  • A term composed of two other terms is fairly straightforward (and its logical implications are too).
  • A term with a complex axiomatization may have components that themselves should be precomposed, which may take the pressure off of need for precomposing the term itself.
• Calculate combinatorial explosion - e.g. meal type x cuisine by country/region = explosion
• Theoretical vs what a common person could recognize as a concept or entity.

Note that a precomposed term therefore has an ontology design pattern or axiomatization (of "necessary and sufficient parts") which, if expressed as an equivalency, yields a way to recognize when the term can apply in or be substituted into the comparable knowledge graph structure.

Its easy to see precomposed terms within a domain, e.g. process or anatomy, but crossing between domains may not make sense:

• cooked x meal type sounds like odd precomposition
• anatomy x process can be odd (except for dried beans, canned beans !!!!)