Address "semantic conflict" for GeneratedBy

[yarikoptic]

We already have GeneratedBy field in BIDS: see e.g. https://bids-specification.readthedocs.io/en/stable/glossary.html#generatedby-metadata . It is a simplified / lightweight version of a more complete GeneratedBy which is being proposed as part of the BEP028. For an example of it look at "Alternative representation for file-level provenance JSON-LD" section.

In

• Allow for GeneratedBy in any sidecar .json file to collect their provenance bids-specification#1970 (comment)

@satra (attn @cmaumet) points to "semantic conflict" between ad-hoc GeneratedBy in BIDS and formalized in BEP028 PROV record. Ideally there should be no such conflicts. I see two potential ways to address

• rename BIDS's GeneratedBy into e.g. BIDSGeneratedBy (suggest better alternatives) and leave it as a simplified
  • worth explicitly listing/stating what would be missing to seamlessly convert to proper PROV GeneratedBy and potentially extend schema to include those fields
• harmonize GeneratedBy to the PROV version.
  • question is either it is not too detailed/cumbersome and simplified version has merit to exist

[satra]

adding an older discussion in relation to this: bids-standard/bids-specification#440 (comment)

in prov there is a property (wasAttributedTo) that connects entities to agents. it's not generatedBy, which connects entities to activities. hence the mismatch in semantics.

[yarikoptic]

I think we are "ok" in that sense of the semantic: there is "Description" field reserved which would/could describe the activity, as I have just done in my adjustment to the

• Allow for GeneratedBy in any sidecar .json file to collect their provenance bids-specification#1970 (comment)

example. And that "Description" aligns well in semantic of BIDS having now those _desc entities and descriptions.tsv - https://bids-specification.readthedocs.io/en/stable/derivatives/common-data-types.html#descriptionstsv , as description of activity lead to the derivative. So, altogether GeneratedBy ATM links to the composite of the agent + activity in a single flattened record.