[Crm-sig] Activity Partitioning (was Actors acting for other Actors)

Robert Sanderson RSanderson at getty.edu
Sat Apr 18 00:24:41 EEST 2020


Oh, I am definitely dubious of the potential about using AI over cultural heritage data for even inferencing based on axioms like this. I am entirely unsurprised that the effort required was unrealistically high. This is why, in the discussion in the Linked Art space, we put it aside as being a theoretical problem, but not one that would cause any real errors.

I think you expressed it perfectly – it is reasonable to assume that properties hold from the whole to the part unless it is otherwise stated. If the whole has a timespan, then the part can be assumed to have that timespan as well (even if an edge case allows it to not in fact have that timespan). With cultural heritage everything is uncertain to a large degree, compared to physical sciences and especially to mathematics, and expecting all assertions to be verified as true is simply impossible.

That said … if there was a sub-property of P9 that was used for activities, we could be more explicit in the scope notes about some of these implications about the actors, rather than just the time and space. If there’s interest in pursuing this, I’m happy to participate and channel use cases. If there isn’t, I’m just as happy to leave the sleeping dog alone 😊

Rob


From: Crm-sig <crm-sig-bounces at ics.forth.gr> on behalf of Martin Doerr <martin at ics.forth.gr>
Date: Friday, April 17, 2020 at 10:15 AM
To: "crm-sig at ics.forth.gr" <crm-sig at ics.forth.gr>
Subject: Re: [Crm-sig] Activity Partitioning (was Actors acting for other Actors)

Dear Robert,

Very good remarks! but...
I may say that this discussion runs headlong against the wall of providing enough information from the real world in order to feed artificial intelligence - you seem yourself to be critical about it.

I can make a qualified statement about that, because I assigned a whole Master thesis about reasoning from parts to wholes in activities, and teams and instrumentation used in activities to a student.  In the framework of the European Project 3D-COFORM, about digitization and creation of 3D models, we could show that even when you intend to monitor completely manually what is going on in a technical process, the effort becomes unrealistically high. If you are interested, I can make the whole thesis available.

Therefore we need inferences that provide reasonable likelihoods: " if there is the activity of writing a book which was carried out by a Person, I don’t think it is legitimate to conclude that the part of writing a chapter was also carried out by that same person." Correct, but it is most reasonable to assume in absence of other evidence. Indeed, historical information almost exclusively of such kind of reasoning, and a large part of empirical natural science as well.

Extending inferences in binary logic with inferences  of likelihood is in my eyes the challenge of the future, and not attempting to model the world until binary logic can deal with it completely. Likelihoods of such inferences can, in enough cases, be approximated by actual distributions. For instance, in a certain context, you may be able to estimate how many writers let chapters write by other people. This is a research agenda I share with other computer scientists.

Please also note, that E4 Period continues to be IsA STV.

Please also note, that the formalization of the CRM does NOT take inverse properties to support different inferences from forward ones, and always imply each other. Hence, using P9i does not make any difference in CRM logic.

All the best,

Martin

On 4/17/2020 6:53 PM, Robert Sanderson wrote:

Dear all,

This discussion (and the partitioning aspect of it) reminds me of a niggling concern that came up in the Linked Art work about scope note of P9 when applied to activities.

In particular, P9 only talks about the part being a subset of the phenomena of the whole:

This property associates an instance of E4 Period with another instance of E4 Period that is defined by a subset of the phenomena that define the former.

To what extent can we infer knowledge through the P9 link, if any? For example, if there is the activity of writing a book which was carried out by a Person, I don’t think it is legitimate to conclude that the part of writing a chapter was also carried out by that same person.

If X consists_of Y, and X carried_out_by Z, then it is not necessarily the case that Y carried_out_by Z, due to the open world assumption.  It could be that X was also carried out by A, B and C, but that was just not stated. Therefore Y could have been carried out by anyone.   And the same argument for all other relationships and properties.

Do we even know that the part is within the same temporal period as the whole? I don’t think so, given that P4 allows alternative opinions about it expressed by assigning multiple Time-Spans to the same E2 Temporal Entity, rather than creating a new E2 and having a 1:1 relationship with TimeSpan. So the part could occur temporally within an undocumented alternative opinion about the timespan. We would thus instead need to also assert P117 occurs during … which is not a sub-property of P9 or vice versa.

P9 is a sub property of P10, which has a domain and range of Spacetime Volume… so this will need to change with the change of STVs no longer being a parent class of Period? At which point we could ensure that P9 implies both P117 and some spatial equivalent?

Conversely, it seems that P9i forms part of IS a strong assertion. If we assert that the part was carried out by A, then the whole MUST have been carried out by at least A, because the carrying-out of the part is a subset of the carrying-out of the whole.  Thus, we should prefer to use P9i, as it enables stronger inferences and understanding.  But … then if we assert that an Activity is part of a Period (rather than merely occurs during it), then the carrying-out-ness is a part of the phenomena of the Period … which cannot be carried out as it’s not an activity.

Result:    :head-exploding-emoji:

For now we have chosen to ignore these issues in linked art for the sake of sanity and convenience. However if there is guidance or improvements that can be made, we would be happy to contribute to those discussions! 😊

Rob

The original issue: https://github.com/linked-art/linked.art/issues/316








--
Dr. Robert Sanderson,  Semantic Architect  |  Getty Digital  |  getty.edu<http://getty.edu/>
[signature_1056976797]



_______________________________________________

Crm-sig mailing list

Crm-sig at ics.forth.gr<mailto:Crm-sig at ics.forth.gr>

http://lists.ics.forth.gr/mailman/listinfo/crm-sig



--

------------------------------------

 Dr. Martin Doerr



 Honorary Head of the

 Center for Cultural Informatics



 Information Systems Laboratory

 Institute of Computer Science

 Foundation for Research and Technology - Hellas (FORTH)



 N.Plastira 100, Vassilika Vouton,

 GR70013 Heraklion,Crete,Greece



 Vox:+30(2810)391625

 Email: martin at ics.forth.gr<mailto:martin at ics.forth.gr>

 Web-site: http://www.ics.forth.gr/isl

CAUTION: This email originated from outside of the Getty. Do not click links or open attachments unless you verify the sender and know the content is safe.


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ics.forth.gr/pipermail/crm-sig/attachments/20200417/d92596e2/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.jpg
Type: image/jpeg
Size: 8036 bytes
Desc: image001.jpg
URL: <http://lists.ics.forth.gr/pipermail/crm-sig/attachments/20200417/d92596e2/attachment-0001.jpg>


More information about the Crm-sig mailing list