Modelling the evolution of descriptions over time
Hope, Aaron (MPBSD)
Greetings,
I wonder if I can ask the group a question about modelling changes to descriptions over time. At the Archives of Ontario we are at the very early stages of planning for a RiC/Linked Open Data proof of concept and will soon be compiling a list of functional requirements for our data model. In recent years we have become more conscious of the need to preserve our descriptions as records in their own right and also as vital context for documenting how the records they describe have been understood and accessed. I am therefore wondering if any of those working on RiC data models have incorporated methods of capturing historical versions of archival descriptions in their graphs, and if so, how they have done so.
At present, the only institution I am aware of which has considered this issue is the UK National Archives with their Project Omega data model, but that is based on the Matterhorn RDF data model instead of RiC-O. They identify at least two potentially viable ways of dealing with changes to description and arrangement over time: 1) creating new entities whenever a description is revised, and 2) retaining only the current state in the RDF graph but preserving past versions in a separate version control system.
I’d be grateful for any thoughts on this topic.
Thank you,
Aaron Hope
Aaron Hope, PhD | Senior Archivist | Archives of Ontario
134 Ian Macdonald Blvd. | Toronto ON M7A 2C5 |
aaron...@ontario.ca | www.archives.gov.on.ca
Richard Williamson
As you have not received a reply yet, I will try to offer a few thoughts, as this is something we in Norway have spent some time experimenting with/thinking about. Firstly, I fully agree with you that this 'meta-metadata' as Project Omega calls it here (item 7)...
https://blog.nationalarchives.gov.uk/digital-archiving-seven-pillars-metadata/
... is very important, and will likely be something that many will wish to express.
Secondly, if you have not already seen it, chapter 6 of the RiC conceptual model...
https://www.ica.org/en/records-in-contexts-conceptual-model
... addresses (at least as I read it) this to at least a certain extent. Indeed, I derive from this chapter the idea that record descriptions should themselves be regarded as, and modelled as, record resources: see Figure 6 for example in the conceptual model, where a description is explicitly modelled (schematically) in this way.
I am keen on this idea: it would allow one to apply the full possibilities to meta-metadata! Taking the self-referentiality a touch further, what I more precisely would like to be able to do is to take an RDF triple 'in RiC', i.e. where the subject, predicate, and object are all defined in RiC, and be able to describe the metadata of such a triple: edits and updates to it, who created it, when, what other triples it relates to, etc. More specifically, I would like to regard such a triple as a record resource in the sense of RiC-O. This kind of self-referentiality is fortunately something that formal semantics has long considered: it is known as 'reification'. Unfortunately, though, reification is not as far developed/fully supported as one might wish in the RDF/OWL world, and as of yet I have not found a solution that I find fully satisfactory. Florence and I discuss this at some length here (see mainly from my second comment on the 8th of May onwards):
https://github.com/ICA-EGAD/RiC-O/issues/56
A number of possibilities are mentioned there: e.g. rdf:Statement is the most 'standard' way in RDF to express reification, but it is not part of OWL. One could just accept this and work in RDF, but it means that one cannot e.g. run an OWL DL reasoner, and after all RiC-O is written in OWL, for good reasons. As Florence mentions in the thread, there are some newer technologies, e.g. RDF-Star (which is worth reading it just for more on reification...
https://www.ontotext.com/knowledgehub/fundamentals/what-is-rdf-star/
...) which might in the end solve this, but support for them in software libraries is not quite there yet, though coming. And whilst I definitely like the idea of RDF-Star, I think it mainly helps mitigate the verbosity of rdf:Statement, rather than solve the issue I mentioned of wishing to use some OWL-native notion of reification; there is something called OWL-Star as well, but it seems quite early-stage.
With all that in mind, one can try to do something directly in RiC/OWL, and I sketched a couple of possibilities in the thread with Florence. They do not fully solve the problem either, but we have tested it in practice, and one can do something at least.
But to summarise, I think that a canonical way to do what you would ask about purely within RiC is likely not yet possible, though as you can see from the thread with Florence, the developers of RiC-O are open to this in the future. What to do in the meantime is an interesting question; if one wishes to stay within RiC and keep things reasonably simple, it might be best to just create a new record resource for each piece of meta-metadata with a free-text R-A16 'Descriptive note', and use e.g. RiC-R022 'is record resource associated with record resource' to link it to the record set which the original metadata is about. E.g. if you change hasCreator for a given record set A from agent B to agent C, one might create a record resource (or some more specific entity) with a descriptive note 'Changed hasCreator from B to C', and use RiC-R022 to relate this new record resource to A. One might alternatively, or in addition, model certain kinds of meta-metadata operations (e.g. 'correcting' or 'changing' some piece of metadata) as activities, and then R033 'documents' might be relevant. If RiC later offers canonical possibilities for more formally/rigorously viewing 'RiC triples' as entities within RiC, one could then replace use of free-text and RiC-R022 with the appropriate things.
I hope there was something useful here! I would be very happy to hear of any ideas/thoughts/suggestions/reflections that others have around this.
Best wishes,
Richard
--
You received this message because you are subscribed to the Google Groups "Records_in_Contexts_users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to Records_in_Context...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/Records_in_Contexts_users/YT3PR01MB8162E7BDCC1FD498CED7C947F9EFA%40YT3PR01MB8162.CANPRD01.PROD.OUTLOOK.COM.
Hope, Aaron (MPBSD)
Hi Richard,
Many thanks for your helpful response, which I certainly found thought provoking. I’m grateful that you drew my attention to the exchange between yourself and Florence Clavaud, and I should have remembered the details and examples in chapter 6 of RiC-CM. I’ll have to give the various options more thought before we decide on the best approach in the interim. I’m also aware that the same principles will come into play when we consider how to document metadata from different sources contributed at different times, e.g. user-contributed descriptions or even corrections to our existing descriptions.
Thanks again,
Aaron
From: Richard Williamson <richard.william...@gmail.com>
Sent: September 7, 2023 6:51 PM
To: Records_in_C...@googlegroups.com; Hope, Aaron (MPBSD) <Aaron...@ontario.ca>
Subject: Re: [Records in Contexts users] Modelling the evolution of descriptions over time
CAUTION -- EXTERNAL E-MAIL - Do not click links or open attachments unless you recognize the sender.