Open dosumis opened 4 years ago
CC @cmungall @balhoff @sbello
Having thought about this some more, I can see potential logical problems with this proposal.
'absent X' will inherit the axiom has_part some ('has fewer parts of type' and (inheres_in some %s) and (towards some X))
That existential quantifier asserts the existence of X. This would not be compatible with the full schema in https://arxiv.org/pdf/1410.3862.pd in which towards is used to infer 'implies presence of'. It may also cause other problems - more investigation and discussion needed. The alternative - using only value restrictions, will lead to a much reduced classification for reduced number classes.
Action item:
STATUS: Draft proposal
For background see https://docs.google.com/presentation/d/1-FMyAdfQhEyDClfKdDHfeC2LZ5Q-dsP5uY_25rEQ-rI/edit#slide=id.g96675c1a66_0_50
https://arxiv.org/pdf/1410.3862.pdf
Challenge: This approach requires punning, which in turn requires declaration of classes as individuals. Do we need to specify these declarations? Can we rely on DOSDP-tools to deal with this automatically? @balhoff ?
lacks/absence
labels are a challenge:
location: dentition part: tooth
=> absent teeth
but
location: head part: hair
=> absent hair on head.
Poss solutions:
reduced number
We need two EC axioms - one with existential restriction in order to => classification by part; one with value restriction to => classification over lacks.