OWL equivalence axioms for symbiosis terms are frequently too open

[cmungall]

here is a subset of the symbiosis terms, their text definitions, their logical definitions, and problems

class	text def	equiv expression	issue
symbiosis	An interspecies interaction between two or more organisms in intimate association	'interspecies interaction between organisms' and (('has occurrent part' some 'multi-organism behavior') and (has_participant min 2 organism))	owl def does not capture the intimate association
antagonism	An ectosymbiosis that results in harm to all participating organisms.	symbiosis and (('has occurrent part' some 'bodily harm') and (has_participant min 2 organism))	satisfied when there is only a single bodily harm individual
ectosymbiosis	A symbiosis where the organisms that are interacting remain physically external to one another.	symbiosis and (has_participant some (organism and ('spatially disjoint from' some organism)))	trivially satisfied as there is always some external organism (coreference problem)
ectosymbiotic commensalism	An ectosymbiosis that results in benefit to one participating organism and neither harm nor benefit to the other participating organisms.	ectosymbiosis and (not (realizes some (homeostasis and ('characteristic of' some organism)))) and ('has part' some 'bodily benefit')	homeostasis not mentioned in text def;"neither harm nor benefit" clause not captured
ectosymbiotic mutualism	An ectosymbiosis that results in benefit to all participating organisms.	ectosymbiosis and ('has part' some 'bodily benefit') and ('has part' only 'bodily benefit')	does not mirror sibling owl definition;combines transitive and universal (bodily benefit is an atom)
obligatory ectosymbiotic commensalism	An ectosymbiotic commensalism that realizes a biological function that inheres in some proper part of the commensal organism that benefits.	'ectosymbiotic commensalism' and (realizes some (function and (('proper part of continuant at all times' some organism) and ('characteristic of' some 'anatomical structure'))))	non-standard use of "obligatory"
endosymbiosis	A symbiosis where one organism is physically contained within another organism (called the 'host').	symbiosis and (has_participant some (organism and ('located in' some organism)))	coreference issue: is the interspecies relationships between two adjacent microbes in a human gut endosymbiosis?
endosymbiotic parasitism	An endosymbiosis that results in benefit to the parasite and harm to the host.	endosymbiosis and ('has part' some ('bodily harm' and (not ('has part' some 'life-death temporal boundary'))))	text and logic deviate (did you mean to include "non-fatal"?)
initimate agonism	An agonism where the participating organisms, although physically external to one another, nevertheless have close and persistent physical contact for most or all of the interaction.	'intimate agonism' and ('has part' some ('bodily harm' and ('has part' some 'life-death temporal boundary')))	circular
transient agonism	An agonism where the participating organisms maintain significant physical separation from one another through most or all of the interaction.	agonism and (has_participant min 2 (organism and ('colocalizes with' some organism)))	text and owl differs (owl uses a poor RO relation I am responsible for)

I would be happy to go into each of these in detail, but overall I suspect you do not derive any benefit from these OWL definitions (i.e if you removed them entailments would be effectively unchanged), and they could be removed (the text definitions are excellent and should remain, that's really all you need).

Some general principles I recommend:

• text definitions and logical definitions should match
• OWL definitions should be kept simple, such that they can be easily reasoned over by humans
• often equivalence axioms can be weakened to subClassOf axioms with no functional loss
• it is better to be conservative when adding equivalence axioms

Following these would involve simply dropping the OWL definitions

I note that in many of the OWL definitions you are hitting the coreference problem. If you are interested I can point you at some papers where the OWL community attempted to solve this by extending OWL2 with a concept called Description Graphs. But ultimately this effort was a bit pointless.

I'd also recommend keeping the text definitions something that could be understood by an ecologist. This is generally followed except in the cases where you veer into BFO language (inheres, realizing, continuant).

[hoganwr]

Thanks Chris! We are undertaking a migration of Apollo-SV to the ODK and will be reviewing these and other more "complicated" axioms. See the pull request that Nico created here: #227

I'd say the resolution of this issue is pending the ODK migration, after which we can review to see what still needs to be fixed. At least with respect to the axioms.

Making definitions more expert friendly is indeed often in tension with ontological precision. In Apollo-SV, we actually used a separate annotation for the expert-friendly version (I think we call the annotation "elucidation" although we were never quite comfortable with it. And I'm not sure we managed to have both for every class that had one or the other).