What relations should be used for informational entities that mirror physical ones

[cmungall]

IAO uses relationships like part-of and has-part to model the structure of documents (informational entities, GDCs). These are broad RO relations, unconstrained by domain and range (but additional universal axioms prevent inappropriate cross upper level category crossings). We seem to have no problem mapping mereological axioms to informational entities to the extent that this doesn't really need any documentation, it's all pretty unintuitive and (relatively) uncontroversial. Perhaps this is because we have a strong mental model of mapping IAO:documents to physical documents.

We could force IAO to say things like "an author summary statement is a concretization of an SDC that inheres in a physical printed piece of text or arrangement of pixels on a physical screen that is part of another piece of text that is the bearer of an SDC that is concretized by an IA:document", as this would be both verbose and probably not even true.

Could this metaphorical projection be extended to other relationships and their mapping to (possibly imaginary) physical structures.

• other mereotopological relationships within a document
• connections between stations on the London tube map
• weights between nodes in an artificial neural network
• adjacency between boxes on a mercator projections of the Earth

I think there are enormous practical problems with broadening the D/Rs of existing relations like connected_to (at best we would need unions, and the addition of universal axioms to prevent cross-category relationships). The implicit advice in the above scenarios would seem to be to create shadow relations, following the same DPs as the material relations.

Assuming this is the case, do we keep allowing the use of partonomic relations for IAO, or switch these out for shadow information-part-of? Arguably part-of is a bit of an outlier already being so generic, and having resisted attempts to push-down to occurrent-part-of style relations (for reasons described in the docs).

I am not arguing against the status quo, it would be good to explicitly codify this rather than have it be implicit and then have things drift off as different ontologies make different interpretations.

[bpeters42]

I was naively thinking that 'has part' / 'part of' for an ICE was unconnected to any physical concretizations. Different 'document parts', like the introduction to a journal article, can be manifested in a myriad of different ways. I thought the OBO relations simply indicate parthood between ICEs with the standard transitivity assumptions, and maybe a property chain saying that if an ICE X has a part Y that 'is about' Z, then X 'is about' Z. What we have been using parthood for is more to say that e.g. a gene expression dataset has a bunch of parts, each of which gives the expression of a single gene. These can be in a table or a barchart or whatever. In practice, we have been trying to avoid talking about the specific concretizations of ICEs, as that doesn't matter for most practical applications. I would prefer to stick to that, and not introduce any RO relations other than 'part-of' to ICEs. Not sure how helpful this is. - Bjoern

…

On Wed, Jul 31, 2024 at 9:06 AM Chris Mungall ***@***.***> wrote: IAO uses relationships like part-of and has-part to model the structure of documents (informational entities, GDCs). These are broad RO relations, unconstrained by domain and range (but additional universal axioms prevent inappropriate cross upper level category crossings). We seem to have no problem mapping mereological axioms to informational entities to the extent that this doesn't really need any documentation, it's all pretty unintuitive and (relatively) uncontroversial. Perhaps this is because we have a strong mental model of mapping IAO:documents to physical documents. We could force IAO to say things like "an author summary statement is a concretization of an SDC that inheres in a physical printed piece of text or arrangement of pixels on a physical screen that is part of another piece of text that is the bearer of an SDC that is concretized by an IA:document", as this would be both verbose and probably not even true. Could this metaphorical projection be extended to other relationships and their mapping to (possibly imaginary) physical structures. - other mereotopological relationships within a document - connections between stations on the London tube map - weights between nodes in an artificial neural network - adjacency between boxes on a mercator projections of the Earth I think there are enormous *practical* problems with broadening the D/Rs of existing relations like connected_to (at best we would need unions, and the addition of universal axioms to prevent cross-category relationships). The implicit advice in the above scenarios would seem to be to create shadow relations, following the same DPs as the material relations. Assuming this is the case, do we keep allowing the use of partonomic relations for IAO, or switch these out for shadow information-part-of? Arguably part-of is a bit of an outlier already being so generic, and having resisted attempts to push-down to occurrent-part-of style relations (for reasons described in the docs <https://oborel.github.io/obo-relations/process-relations/#process-part-wholes> ). I am not arguing against the status quo, it would be good to explicitly codify this rather than have it be implicit and then have things drift off as different ontologies make different interpretations. — Reply to this email directly, view it on GitHub <#802>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADJX2ISGZOQT6JBWCLAZOKDZPEDPPAVCNFSM6AAAAABLYZKTAGVHI2DSMVQWIX3LMV43ASLTON2WKOZSGQ2DANBQHE3TANA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

-- Bjoern Peters Professor La Jolla Institute for Immunology 9420 Athena Circle La Jolla, CA 92037, USA Tel: 858/752-6914 Fax: 858/752-6987 http://www.liai.org/pages/faculty-peters