W3C home > Mailing lists > Public > public-esw-thes@w3.org > April 2014

RE: TGN place types (broader/narrower spanning ConceptSchemes)

From: Vladimir Alexiev <vladimir.alexiev@ontotext.com>
Date: Tue, 1 Apr 2014 23:07:19 +0300
To: <public-esw-thes@w3.org>
Cc: <jcobb@getty.edu>
Message-ID: <038501cf4de5$fcddfd50$f699f7f0$@alexiev@ontotext.com>
Hi folks, thanks for the substantial comments!!
But the discussion strayed quite far from my original question.
Please give me more specific answers to my question...


Osma Suominen> How about just declaring the place types as rdfs:subClassOf
> skos:Concept, and then use rdf:type to represent the relation between
> places and place types?
> For publishing purposes, a SKOS representation is derived, simply by 
> defining rdfs:subClassOf and rdfs:subPropertyOf relationships that map 
> the classes and properties of the ontology to a simplified SKOS view

I think this is a bad idea:
- Place Types come from a thesaurus that is refreshed every 2 weeks. I don't think people appreciate ontologies that may change every 2 weeks
- The Place Type relation carries historic info, whereas nobody would expect rdf:type to be time-dependent

> Though the fact that the place types are also AAT concepts might be a
> problem here, because then the place types must be both skos:Concept and
> owl:Class instances at the same time. 

This too.


Oreste Signore> This represents the present status
> tgn:7018759-concept a skos:Concept; # "Sofiya-Grad";
>  gvp:broaderPartitive tgn:7006413-concept. # Bulgaria
> What about Crimea today or one month ago? (just to recall very recent events).
> In other words, we are loosing the "time dimension"

We are not. 
Hierarchical relations in TGN (just like in AAT) carry historic information, which is attached with rdf:Statement.
See http://www.getty.edu/research/tools/vocabularies/lod/aat_semantic_representation.pdf, sec 2.14 and 2.14.2

It's not a fully-fledged account of historic names like I've seen in Pleiades/Pelagios, but such account is not in scope of TGN.

> the paper: http://www.weblab.isti.cnr.it/papers/public/ch2013-Athens-Signore-paper.pdf
> the slides: http://www.weblab.isti.cnr.it/papers/public/ch2013-Athens-Signore-slides.pdf

Thanks, I'll read them.
But my question was about Place Types (which also are time-dependent), not super-places.


Richard Light> In your paper, you say early on that  
> Place types have Historic Info: Historic flag, Start Date, End Date, and Comment (display date).
> Surely that's wrong: the types don't of themselves have these properties

I think you missed the previous sentence: "When applied to a particular place"

> it follows that the relationship between a place and a specified place type can't be a simple "has type" property, 
> linking place and place type. 
> If you do that, there is nowhere to hang these additional place type properties (a classic problem with simplistic RDF).

Hierarchical relations in TGN (just like in AAT) carry historic information, which is attached with rdf:Statement.
If we model the "place type relation" with gvp:broaderInstantial (as I proposed), we'll use exactly the same mechanism

> I don't see how the "broader" relationships between place types impact on the "broader" relationships between the place instances themselves

- place types (like most other AAT concepts) are related with gvp:broaderGeneric
- places (in TGN) will be related with gvp:broaderPartitive
- my question is: shouldn't then I use gvp:broaderInstantial (from TGN to AAT) for the "place type relation"

> the relationship could be defined in the context of something like a CIDOC CRM E4 Period

I know about E13 Attribute Assignment and its sub-class E15 Identifier Assignment.
Including the fact that E13 is missing the analog of rdf:predicate, so for the British Museum we had to make an extension sub-class bmo:EX_Association with a prop called bmo:PX_property.
But that's not my question.

> What would be really helpful for end-users would be if the various historically-aware place authorities
> (SAPO, TGN, Pelagios, PastPlace, ...) all used the same ontological structure
> to express statements about places in time.  Possibly something CIDOC CRM-based

I agree, that would be useful!
But I think TGN should follow not lead here, since it doesn't have a full account of historic place names like Pelagios
(and from the name, I guess PastPlace is mostly about that)

CRM can probably represent this, however we saw a problem while working on the British Museum representation:

1. There are specified Events for the beginning/end of things, eg
- Person: Birth/Death
- Thing: Production/Destruction
- Identifier: Identifier Assignment/deassignment
- Place Appellation: Attribute Assignment (no deassignment, unfortunately)

2. But there are no specified Periods for the same things:
- Person: life, floruit (when/where was active), reign, ...
- Thing: existence
- Identifier: when/where was used

We ended up using "P12i was present at" for 2, but that is very unspecific.


Christophe Dupriez> I would not try to add all the modelling data in SKOS: I would rely on another ontology for this...
> use a SPARQL queries to create automatically and periodically 
> SKOS graphs starting from ontological information served by a (external) SPARQL server

So you propose to derive an instant-in-time SKOS snapshot of time-varying data representing other ontologies? I think this is a bad idea:
- Thesauri do not hold only current info, they also need to hold historic info
- Thesauri are used for findability (e.g. in documentary retrieval engines as you wrote).
  If we have a thesaurus that says "Crimea after 1954 was part of Ukraine, but after 2014 is part of Russia", how would we use that to index documents?
  How would the indexing engine figure out whether a book talks about Crimea before 1954, after 1954, or after 2014? 
  What if the book talks about all of these periods?

I think that for findability purposes, the thesaurus should have both of these statements:
  <Crimea> broaderPartitive <Ukraine>.
  <Crimea> broaderPartitive <Russia>.

(Some might argue they are not logically consistent...)
Then the rdf:Statement additions can be consulted to figure out which was true at a particular date (if that is interesting)


Bernard Vatant>
> The more I use SKOS, the more I see it that way : a translation of knowledge fit for a certain type of application. 
> Everything (every thing) can be re-presented as a skos:Concept for some purpose, but nothing is a skos:Concept in essentia.

I agree with you, and the plan is to represent TGN places dually, see the last section of my paper:
"It's clear that TGN will be published as a SKOS thesaurus, with additional place-specific classes and properties. We'll respect a "concept-thing" dichotomy, using foaf:focus to relate the two"

But coming back to my question:

> In such a configuration you get a hierarchy of "place-concepts" with an orthogonal facet of "type-concepts",
> both in SKOS, and it works nicely for faceted classification.

So... Do you think it's a good idea to use gvp:broaderPartitive for the former, and gvp:broaderInstantial (spanning TGN->AAT) for the latter?


Rimvydas Laužikas>
> the main problem is "preferred type", actually, preferences is not "reality object", but our "mind construction"

Plaese take a look at 
https://www.getty.edu/research/tools/vocabularies/tgn_in_depth.pdf  p29
(It's a good example for everyone to look at, since it shows historic info on place types)

It gives the place types of Indianapolis:
inhabited place ...  founded in 1821 (Start=1821)
state capital ...  since 1825 (Start=1825)
manufacturing center
transportation center
sporting center ...  especially noted for Indianapolis 500 automobile race, since 1911 (Start=1911)
financial center

I think most people will agree that "inhabited place" and "state capital" are generally more important than "sporting center".
Anyway, Getty's preferred type ("inhabited place") is used to display in the hierarchy, and certainly I must emit it somehow.

Even CRM has "P48 has preferred identifier". 
BTW, I think this is not so useful... "Preferred Title" and "Preferred Image Representation" are needed more:

> Place type child can be composed from 2 or more (typological, chronological, geographical and etc.) parents

So.. do you like or dislike my proposal?


Walter Koch
> might also be useful to harmonize vocabularies, e.g. placetypes (categories?)

After comparing Pleiades and TGN, I think this would be quite an ambitious undertaking.
And let's not forget OpenStreetMap features, all the way down to phone boxes and trash bins :-)

> http://demo.ait.co.at/thesaurus/index.php?file=xml/Operation_scanIndex_PLACETYPE.xml
> http://demo.ait.co.at/thesaurus/index.php?file=xml/complexQuery_rivers_in_germany.xml

Hey, that's a very nice source of "competence questions" that our representation should be able to answer with SPARQL queries in a natural way.
I'll make sure to include similar sample queries 
(see last sections of http://www.getty.edu/research/tools/vocabularies/lod/aat_semantic_representation.pdf for AAT sample queries).

(BTW, this web service might be contradicting some Getty license terms, but now that they're opening everything as LOD, I don't think they'll mind terribly)

It also shows the richness of TGN data, in XML form.
Received on Tuesday, 1 April 2014 20:07:45 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 19:46:36 UTC