Compare how CRM is used in ARIADNE triple store

[nichtich]

See recent paper https://doi.org/10.11141/ia.64.12 and RDF data in CRM at http://openarchaeo.huma-num.fr/

[nichtich]

For dates it looks like they use crm:E52_Time-Span as class with crm:P82a_begin_of_the_begin and crm:P82b_end_of_the_end, and skos:prefLabel, for instance:

<https://iceramm.huma-num.fr/bdresultats.php?notice=485/metadata>
  crm:P4_has_time-span <https://iceramm.huma-num.fr/bdresultats.php?notice=485/metadata/datation> .
<https://iceramm.huma-num.fr/bdresultats.php?notice=485/metadata/datation>
  skos:prefLabel "2018-02-06" ;
  crm:P82a_begin_of_the_begin "2018-02-06T00:00:00"^^xsd:dateTime ;
  crm:P82b_end_of_the_end "2018-02-06T23:59:59"^^xsd:dateTime .

Could be simplified to:

<https://iceramm.huma-num.fr/bdresultats.php?notice=485/metadata>
  crm:P4_has_time-span "2018-02-06"^^xsd:date

Sure there are cases where more precise start and end is needed, but in this case it's just a date and the exact timestamps are useless without time zone anyway.

[larsgw]

CIDOC does recommend using P82a/P82b in their guidance document (https://www.cidoc-crm.org/sites/default/files/issue%20443%20-%20Implementing%20CIDOC%20CRM%20in%20RDF%20v1.1.pdf).

[nichtich]

Given that P82a/P82b is not part of CIDOC-CRM 7.1.3 (released on February 2024), although recommended in October 2020 does not give me confidence, but we may have to live with it. Unless the time in dateTime instances differ from 00:00:00 and 23:59:59, respectively, they seem to give a wrong impression of accuracy. I strongly believe the original data is just a date, so using P82a/P82b is denormalization better done by explicit inference rules. Anyway, ARIADNE uses P82a/P82b so we have to live with it.