VEP-015: relationship_type#References

Tue Mar 5 18:59:54 CET 2024

I think this is a really interesting discussion and being able to define relationships between resources/datasets is something that we are very interested in as well.

It seems to me that there are all different kinds of ways that you could identify (potentially) related resources and this “reference paper of X cites reference paper of Y” relationship is just one particular inference that could be made. You could potentially make an inference based on some kind of similarity score or reference paper author overlap or many different other things. The benefit of having a very specific relationship is more clarity for the end user and making it easier to apply machine-based techniques. E.g. if I have a very specific citation relationship then I could build a graph of all the resources based on it and do some other reasoning. But also, a more generic “relatesTo” relationship could be used in many different ways, because it might not be practical to add specific vocabulary terms for each type of relationship one might want to explore.

Unfortunately, I don’t really have any concrete ideas of the best way to do this. I can see the argument both ways. We want to be able to link resources and tell the users “You are looking at X. You might also want to look at Y because of ____ reason.” So I wonder if something more complicated than a simple relationship might be needed and something more like a relationship plus a provenance is required?

Also, is it a concern that “reference paper of X cites reference paper of Y” wouldn’t apply to all resource types?

-Sarah Weissman (MAST/STScI)

From: semantics <semantics-bounces at ivoa.net> on behalf of Anne Catherine Raugh via semantics <semantics at ivoa.net>
Reply-To: Anne Catherine Raugh <araugh at umd.edu>
Date: Tuesday, March 5, 2024 at 11:16 AM
To: gilles landais <gilles.landais at astro.unistra.fr>
Cc: "registry at ivoa.net" <registry at ivoa.net>, "semantics at ivoa.net" <semantics at ivoa.net>
Subject: Re: VEP-015: relationship_type#References

External Email - Use Caution
I'm confused by the statistics remark - the point of including any of these relationships in the DOI metadata is to generate statistics. Sounds like you're complaining that it works. (Let me know if you want the long, pedantic aside on "Cites" and "References" in DataCite metadata.)

Sticking to the point, it sounds to me like Vizier has defined "references" for their own context, which is pretty much what ends up happening at each institution writing DOI metadata for DataCite.

Rather than propagating ambiguity to the IVOA interface, I would suggest that if we are going to get into the relationship-defining business (and it's hard not to), we should be as specific in the relationship definition as practical. This is especially true if the same concept applies to anything other than Vizier. Vocabulary lists with vague and circular definitions (yes, I have read the DataCite documentation 😒) are in some sense worse than free-text fields, because at least in free-text fields there's a good chance the user is saying what they actually intend, if only there is a human around to read it.

Thinking on a larger scale, it is also problematic to define seemingly general concepts like "references" in a context-specific way. In other words, whatever we do for Vizier, we should avoid a solution that might cause confusion in contexts where "referencing" and "citing" are distinct relationships.

-Anne.

On Tue, Mar 5, 2024 at 9:10 AM gilles landais <gilles.landais at astro.unistra.fr<mailto:gilles.landais at astro.unistra.fr>> wrote:

Hi everyone,

"references" was a VizieR proposal to replace the deprecated ivoa term "related-to" (eg: https://cds.unistra.fr/registry/?verb=GetRecord&metadataPrefix=ivo_vor&identifier=ivo://cds.vizier/J/AJ/161/36<https://urldefense.com/v3/__https:/cds.unistra.fr/registry/?verb=GetRecord&metadataPrefix=ivo_vor&identifier=ivo:**Acds.vizier*J*AJ*161*36__;Ly8vLy8v!!CrWY41Z8OgsX0i-WU-0LuAcUu2o!3LY2l3pX-3lBPmMJGAqV8gnhVYDZgTVKK72TtNZ_KRKUL8jgA6r-loWRy9qHf-cyE7prGMLAaNYAZyNbRjI$>).
VizieR uses "related-to" on a catalogue to list VizieR datasets which was attached to articles cited by the catalogue reference article (the "See also" section in VizieR ReadMe file).

This list of resources is not really a list of citation (but a "citation by rebounds" ...). It is why we used the term "related-to" (and we liked it). That's an option, provided for discoverability reasons.
(Note that this VizieR list is provided only in its VO registry record but not in its DOI metadata)

If you think, "references" is not adapted for this VizieR use case (and according to the DataCite definition, it seems that it is not)  we can remove this list or keep "related-to".

(Note: In fact, I don't like "Cites" because in DataCite API "Cites" relation increases the datacite statistics (curl https://api.datacite.org/dois/<https://urldefense.com/v3/__https:/api.datacite.org/dois/__;!!CrWY41Z8OgsX0i-WU-0LuAcUu2o!3LY2l3pX-3lBPmMJGAqV8gnhVYDZgTVKK72TtNZ_KRKUL8jgA6r-loWRy9qHf-cyE7prGMLAaNYA5oOzV60$>{DOI}))

Thank for your advises,

with best,
Gilles Landais (CDS)

Le 05/03/2024 à 14:15, Henneken, Edwin via registry a écrit :
I agree with Anne. In their Metadata Schema 4.5 document, DataCite uses pretty much the same language for "References" as proposed here

DataCite Definition: indicates B is used as a source of information for A
URL: https://datacite-metadata-schema.readthedocs.io/en/4.5/appendices/appendix-1/relationType/#references<https://urldefense.com/v3/__https:/datacite-metadata-schema.readthedocs.io/en/4.5/appendices/appendix-1/relationType/*references__;Iw!!CrWY41Z8OgsX0i-WU-0LuAcUu2o!3LY2l3pX-3lBPmMJGAqV8gnhVYDZgTVKK72TtNZ_KRKUL8jgA6r-loWRy9qHf-cyE7prGMLAaNYAVSmZDTY$>

So, yes, you could use it for "cited-reference" if you wish.

DataCite provides a nice page on the use of the relationship types Cited, IsCitedBy, References and IsReferenceBy:
https://support.datacite.org/docs/contributing-citations-and-references<https://urldefense.com/v3/__https:/support.datacite.org/docs/contributing-citations-and-references__;!!CrWY41Z8OgsX0i-WU-0LuAcUu2o!3LY2l3pX-3lBPmMJGAqV8gnhVYDZgTVKK72TtNZ_KRKUL8jgA6r-loWRy9qHf-cyE7prGMLAaNYA50mZ4bI$>

But, due to its more general nature, I think using "Cited" is to be preferred over "References" when the relationship is for a "citation-reference".

From the ADS point of view, if we get metadata for a work to be indexed and it has a RelatedIdentifier section, we don't want to be guessing what the precise meaning/role of a specific relationType is; different values for relationType result in different actions within the ADS indexing workflow.

All the best
Edwin

Edwin Henneken (he | him | his)

NASA Astrophysics Data System

Center for Astrophysics | Harvard & Smithsonian

60 Garden Street | MS 83 | Cambridge, MA 02138

Error! Filename not specified.

ads.harvard.edu<https://urldefense.com/v3/__http:/ads.harvard.edu/__;!!CrWY41Z8OgsX0i-WU-0LuAcUu2o!3LY2l3pX-3lBPmMJGAqV8gnhVYDZgTVKK72TtNZ_KRKUL8jgA6r-loWRy9qHf-cyE7prGMLAaNYAvtYC98o$> | @adsabs<https://urldefense.com/v3/__https:/twitter.com/adsabs__;!!CrWY41Z8OgsX0i-WU-0LuAcUu2o!3LY2l3pX-3lBPmMJGAqV8gnhVYDZgTVKK72TtNZ_KRKUL8jgA6r-loWRy9qHf-cyE7prGMLAaNYAGvZ9CnY$> | ui.adsabs.harvard.edu<https://urldefense.com/v3/__https:/ui.adsabs.harvard.edu/__;!!CrWY41Z8OgsX0i-WU-0LuAcUu2o!3LY2l3pX-3lBPmMJGAqV8gnhVYDZgTVKK72TtNZ_KRKUL8jgA6r-loWRy9qHf-cyE7prGMLAaNYAHXLZoXk$>

cfa.harvard.edu<https://urldefense.com/v3/__https:/cfa.harvard.edu/__;!!CrWY41Z8OgsX0i-WU-0LuAcUu2o!3LY2l3pX-3lBPmMJGAqV8gnhVYDZgTVKK72TtNZ_KRKUL8jgA6r-loWRy9qHf-cyE7prGMLAaNYAvp-Nl6E$> | Facebook<https://urldefense.com/v3/__http:/cfa.harvard.edu/facebook__;!!CrWY41Z8OgsX0i-WU-0LuAcUu2o!3LY2l3pX-3lBPmMJGAqV8gnhVYDZgTVKK72TtNZ_KRKUL8jgA6r-loWRy9qHf-cyE7prGMLAaNYAMXVnY8Q$> | Twitter<https://urldefense.com/v3/__http:/cfa.harvard.edu/twitter__;!!CrWY41Z8OgsX0i-WU-0LuAcUu2o!3LY2l3pX-3lBPmMJGAqV8gnhVYDZgTVKK72TtNZ_KRKUL8jgA6r-loWRy9qHf-cyE7prGMLAaNYAZ_c8_xs$> | YouTube<https://urldefense.com/v3/__http:/cfa.harvard.edu/youtube__;!!CrWY41Z8OgsX0i-WU-0LuAcUu2o!3LY2l3pX-3lBPmMJGAqV8gnhVYDZgTVKK72TtNZ_KRKUL8jgA6r-loWRy9qHf-cyE7prGMLAaNYAiF1_vpA$> | Newsletter<https://urldefense.com/v3/__http:/cfa.harvard.edu/newsletter__;!!CrWY41Z8OgsX0i-WU-0LuAcUu2o!3LY2l3pX-3lBPmMJGAqV8gnhVYDZgTVKK72TtNZ_KRKUL8jgA6r-loWRy9qHf-cyE7prGMLAaNYAfIqNCY8$>

On Tue, Mar 5, 2024 at 7:00 AM Anne Catherine Raugh via semantics <semantics at ivoa.net<mailto:semantics at ivoa.net>> wrote:
"References" is actually a problematic term for DataCite, largely because there is also the term "Cites" to confuse the issue. Both are used for citation-reference relationships, depending on the whim of the metadata author. "References" is also used for other relationships when one resources mentions another (in acknowledgements, for example, or as part of a "For further information" reading list).

If the intention is for all relationships of this new type "reference" to be interpreted as a "citation-reference", then you should probably say so explicitly. It seems unlikely that other sorts of relationships would be relevant to the IVOA case, and perhaps it might encourage citation of results where appropriate.

-Anne.

On Tue, Mar 5, 2024 at 3:39 AM Markus Demleitner via semantics <semantics at ivoa.net<mailto:semantics at ivoa.net>> wrote:
Dear Semantics, dear Registry,

We have a new VEP on the Registry's vocabulary of relationship types,
VEP-015: https://github.com/ivoa-std/VEPs/blob/main/VEP-015.txt<https://urldefense.com/v3/__https:/github.com/ivoa-std/VEPs/blob/main/VEP-015.txt__;!!CrWY41Z8OgsX0i-WU-0LuAcUu2o!3LY2l3pX-3lBPmMJGAqV8gnhVYDZgTVKK72TtNZ_KRKUL8jgA6r-loWRy9qHf-cyE7prGMLAaNYAS3qmfwY$>.

Here is its text:

  Vocabulary: http://www.ivoa.net/rdf/voresource/relationship_type<https://urldefense.com/v3/__http:/www.ivoa.net/rdf/voresource/relationship_type__;!!CrWY41Z8OgsX0i-WU-0LuAcUu2o!3LY2l3pX-3lBPmMJGAqV8gnhVYDZgTVKK72TtNZ_KRKUL8jgA6r-loWRy9qHf-cyE7prGMLAaNYArBBlu_0$>
  Author: gilles.landais at unistra.fr<mailto:gilles.landais at unistra.fr>
  Date: 2022-12-15

  New Term: References
  Action: Addition
  Label: references
  Description: This resource used the related resource as a source of information.

  Used-in: The registry record ivo://edu.gavo.org/hd/gavo_addpms<https://urldefense.com/v3/__http:/edu.gavo.org/hd/gavo_addpms__;!!CrWY41Z8OgsX0i-WU-0LuAcUu2o!3LY2l3pX-3lBPmMJGAqV8gnhVYDZgTVKK72TtNZ_KRKUL8jgA6r-loWRy9qHf-cyE7prGMLAaNYAwerL9zs$> (and most
    other IVOA document records; cf.
    <http://dc.zah.uni-heidelberg.de/wirr/q/ui/fixed?field0=restype&operator0=%3D&operand0=doc%3Adocument<https://urldefense.com/v3/__http:/dc.zah.uni-heidelberg.de/wirr/q/ui/fixed?field0=restype&operator0=*3D&operand0=doc*3Adocument__;JSU!!CrWY41Z8OgsX0i-WU-0LuAcUu2o!3LY2l3pX-3lBPmMJGAqV8gnhVYDZgTVKK72TtNZ_KRKUL8jgA6r-loWRy9qHf-cyE7prGMLAaNYAhBkz85c$>>)

  Rationale: This term is proposed to replace the deprecated term "related-to"
    in the registry relationships. In VizieR, "related-to" refers VO resources
    which are linked by their biliographic references.

    For instance, a VizieR catalogue V is attached to a bibliographic reference.
    This article includes in the section "References" citations to other
    articles A1, A2, ... some of them have a VizieR catalogue V1,V2,...
    The proposed relation makes the relation between V and V1,V2,...

    "References"  is a term used in DataCite schema.

Do the WGs have opinions on this?  I'd suggest comments should be
addressed to the Registry mailing list.

Thanks,

          Markus
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.ivoa.net/pipermail/registry/attachments/20240305/3732bd06/attachment-0001.htm>