Re: [Wikidata-l] Questions about statement qualifiers

[Antoine Isaac]

Hi Antoine, all,

I was also a bit puzzled by this. If you want more discussion I there is stuff 
on Gerard's blog [1,2].
After some patient explanations of the kind on this list, I think I understood 
what qualifiers are about.

Still I disagree with a part of what Markus said. Trying to understand claims as 
"statement about statements", as Antoine did, is not being concerned only about 
the informal meaning. It is a rather deep data modeling issue.

In wikidata a qualifier can be about the object of a claim or about the claim itself (Markus' "meta" level), 
and there's no means to distinguish one from the other in the form of the data. In fact such "data structure" 
for qualifiers is much more dependent on an "informal reading" than thought: it fits really well how humans 
would enter and "read" the data, but less well what a machine would need to exploit it (as the construct is 
intrinsically ambiguous on whether it's an amendment to the truth conditions of a claim, or a n-ary relation).

I'm not saying it's bad per se. The discussion made me understand better why it 
was designed so, and I can understand the advantages. Still I am not so sure 
it's really a winner in terms of interoperability with other systems.

Best,

Antoine

http://ultimategerardm.blogspot.nl/2013/10/more-heady-stuff-about-wikidata-and.html
http://ultimategerardm.blogspot.nl/2013/10/abdallah-ii-six-times-sultan-of-morocco.html



Hello Antoine,

just to add to what was already said:

a Qualifier in Wikidata is not a "statement about a statement". In RDF semantics, the pattern 
that we follow is not the reification of the triple and then to make triples with the reified triple as a 
subject, as per  but rather the pattern of n-ary relations 
per  . The use cases very beautifully visualize how 
Wikidata maps to RDF: 

This is also what Wikidata's mapping to RDF document explains and motivates: 


I hope this helps,

Denny



On Oct 31, 2013 3:40 AM, "Antoine Zimmermann" mailto:antoine.zimmerm...@emse.fr>> wrote:

Hello,


I have a few questions about how statement qualifiers should be used.


First, my understanding of qualifiers is that they define statements about 
statements. So, if I have the statement:

Q17(Japan)  P6(head of government)  Q132345(Shinzō Abe)

with the qualifier:

  P39(office held)  Q274948(Prime Minister of Japan)

it means that the statement holds an office, right?
It seems to me that this is incorrect and that this qualifier should in 
fact be a statement about Shinzō Abe. Can you confirm this?



Second, concerning temporal qualifiers: what does it mean that the "start" or "end" 
is "no value"?  I can imagine two interpretations:

  1. the statement is true forever (a person is a dead person from the 
moment of their death till the end of the universe)
  2. (for end date) the statement is still true, we cannot predict when 
it's going to end.

For me, case number 2 should rather be marked as "unknown value" rather than "no 
value". But again, what does "unknown value" means in comparison to having no indicated value?



Third, what if a statement is temporarily true (say, X held office from T1 to 
T2) then becomes false and become true again (like X held same office from T3 to 
T4 with T3 > T2)?  The situation exists for Q35171(Grover Cleveland) who has 
the following statement:

Q35171  P39(position held)  Q11696(President of the United States of 
America)

with qualifiers, and a second occurrence of the same statement with 
different qualifiers. The wikidata user interface makes it clear that there are 
two occurrences of the statement with different qualifiers, but how does the 
wikidata data model allows me to distinguish between these two occurrences?

How do I know that:

  P580(start date)  "March 4 1885"

only applies to the first occurrence of the statement, while:

  P580(start date)  "March 4 1893"

only applies to the second occurrence of the statement?
I could have a heuristic that says if two "start date"s are given, then 
assume that they are the starting points of two disjoint intervales. But can I always 
guarantee this?


Best,
AZ

--
Antoine Zimmermann
ISCOD / LSTI - Institut Henri Fayol
École Nationale Supérieure des Mines de Saint-Étienne
158 cours Fauriel
42023 Saint-Étienne Cedex 2
France
Tél:+33(0)4 77 42 66 03 
Fax:+33(0)4 77 42 66 66 
http://zimmer.__aprilfoolsreview.com/ 

_
Wikidata-l mailing list
Wikidata-l@lists.wikimedia.org 
https://lists.wikimedia.org/__mailman/listinfo/wikidata-l 

Re: [Wikidata-l] Questions about statement qualifiers

[Denny Vrandečić]

Hello Antoine,

just to add to what was already said:

a Qualifier in Wikidata is not a "statement about a statement". In RDF
semantics, the pattern that we follow is not the reification of the triple
and then to make triples with the reified triple as a subject, as per <
http://www.w3.org/TR/rdf-mt/#ReifAndCont> but rather the pattern of n-ary
relations per  . The use cases
very beautifully visualize how Wikidata maps to RDF: <
http://www.w3.org/TR/swbp-n-aryRelations/#useCase1>

This is also what Wikidata's mapping to RDF document explains and
motivates: 

I hope this helps,

Denny



On Oct 31, 2013 3:40 AM, "Antoine Zimmermann" 
wrote:

> Hello,
>
>
> I have a few questions about how statement qualifiers should be used.
>
>
> First, my understanding of qualifiers is that they define statements about
> statements. So, if I have the statement:
>
> Q17(Japan)  P6(head of government)  Q132345(Shinzō Abe)
>
> with the qualifier:
>
>  P39(office held)  Q274948(Prime Minister of Japan)
>
> it means that the statement holds an office, right?
> It seems to me that this is incorrect and that this qualifier should in
> fact be a statement about Shinzō Abe. Can you confirm this?
>
>
>
> Second, concerning temporal qualifiers: what does it mean that the "start"
> or "end" is "no value"?  I can imagine two interpretations:
>
>  1. the statement is true forever (a person is a dead person from the
> moment of their death till the end of the universe)
>  2. (for end date) the statement is still true, we cannot predict when
> it's going to end.
>
> For me, case number 2 should rather be marked as "unknown value" rather
> than "no value". But again, what does "unknown value" means in comparison
> to having no indicated value?
>
>
>
> Third, what if a statement is temporarily true (say, X held office from T1
> to T2) then becomes false and become true again (like X held same office
> from T3 to T4 with T3 > T2)?  The situation exists for Q35171(Grover
> Cleveland) who has the following statement:
>
> Q35171  P39(position held)  Q11696(President of the United States of
> America)
>
> with qualifiers, and a second occurrence of the same statement with
> different qualifiers. The wikidata user interface makes it clear that there
> are two occurrences of the statement with different qualifiers, but how
> does the wikidata data model allows me to distinguish between these two
> occurrences?
>
> How do I know that:
>
>  P580(start date)  "March 4 1885"
>
> only applies to the first occurrence of the statement, while:
>
>  P580(start date)  "March 4 1893"
>
> only applies to the second occurrence of the statement?
> I could have a heuristic that says if two "start date"s are given, then
> assume that they are the starting points of two disjoint intervales. But
> can I always guarantee this?
>
>
> Best,
> AZ
>
> --
> Antoine Zimmermann
> ISCOD / LSTI - Institut Henri Fayol
> École Nationale Supérieure des Mines de Saint-Étienne
> 158 cours Fauriel
> 42023 Saint-Étienne Cedex 2
> France
> Tél:+33(0)4 77 42 66 03
> Fax:+33(0)4 77 42 66 66
> http://zimmer.aprilfoolsreview.com/
>
> ___
> Wikidata-l mailing list
> Wikidata-l@lists.wikimedia.org
> https://lists.wikimedia.org/mailman/listinfo/wikidata-l
>
___
Wikidata-l mailing list
Wikidata-l@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikidata-l

Re: [Wikidata-l] Questions about statement qualifiers

[Markus Krötzsch]

Hi Antoine,

The main answer to your questions is that the data model of Wikidata 
defines a *data structure* not the *informal meaning* that this data 
structure has in an application context (that is: what we, humans, want 
to say when we enter it). I try to explain this a bit better below.


How the presence or absence of a qualifier contributes to the informal 
meaning of a statement is not something that is defined by Wikidata. 
Just like Wikidata does not define what the property "office held" 
means, it also does not define what it means if "office held" is used 
with additional qualifiers. This is entirely governed by the community 
who uses these structures to express something. Of course, the community 
tries to do this in a systematic, reasoned, and intuitive way. However, 
there will never be a general rule how to interpret an arbitrary quantifier.


In particular, it is not true that quantifiers are "statements about 
statements". First, to avoid confusion, I need to explain Wikidata's 
terminology. A "statement" in Wikidata comprises the whole data 
structure: main property and value, qualifiers, and references. The 
structure without the references (the thing that the references provide 
evidence for) is called a "claim". A claim thus contains a main property 
and value (or no value or unknown value) and zero or more qualifier 
properties with values. Every claim encodes something that is claimed 
about the subject of the page (the Wikidata entity), and the references 
given are supporting this claim (as a whole).


You already illustrated yourself how this is different from making 
"statements about statements": it would lead to confusion when several 
statements have the same main property-value but different qualifiers. 
This is also why our RDF export does not use the same resource for 
reifying statements with the same main property-value. Instead, we only 
share the same resource it two claims are completely identical 
(including qualifiers).


It is true that many qualifiers have a certain "meta" flavour, but this 
is not always the case. An interesting case that you might have seen is 
P161 ("cast member") that is used to denote the actors in a film. The 
typical qualifier there is P453 ("role"), used to name the role 
(character) that the person played in the film. If you look at this, 
this is more like a ternary relation hasActor(film,actor,role) than like 
a "meta-statement". Indeed, an n-ary relationship cannot in general be 
represented by a meta-statement about a binary relation, again for the 
same reasons that you gave in your email. In this view, one should maybe 
also think of a relationship usPresident(person,start date, end date) 
than of an annotated assertion usPresident(person). Wikidata is special 
in that qualifiers are optional, yet the modelling view of n-ary 
relations might be closer to the pragmatic truth, since it avoids any 
meta-statements (it also elegantly justifies why there are no 
meta-meta-statements, i.e., qualifiers on qualifiers).


Best regards,

Markus


On 31/10/13 11:39, Antoine Zimmermann wrote:

Hello,


I have a few questions about how statement qualifiers should be used.


First, my understanding of qualifiers is that they define statements
about statements. So, if I have the statement:

Q17(Japan)  P6(head of government)  Q132345(Shinzō Abe)

with the qualifier:

  P39(office held)  Q274948(Prime Minister of Japan)

it means that the statement holds an office, right?
It seems to me that this is incorrect and that this qualifier should in
fact be a statement about Shinzō Abe. Can you confirm this?



Second, concerning temporal qualifiers: what does it mean that the
"start" or "end" is "no value"?  I can imagine two interpretations:

  1. the statement is true forever (a person is a dead person from the
moment of their death till the end of the universe)
  2. (for end date) the statement is still true, we cannot predict when
it's going to end.

For me, case number 2 should rather be marked as "unknown value" rather
than "no value". But again, what does "unknown value" means in
comparison to having no indicated value?



Third, what if a statement is temporarily true (say, X held office from
T1 to T2) then becomes false and become true again (like X held same
office from T3 to T4 with T3 > T2)?  The situation exists for
Q35171(Grover Cleveland) who has the following statement:

Q35171  P39(position held)  Q11696(President of the United States of
America)

with qualifiers, and a second occurrence of the same statement with
different qualifiers. The wikidata user interface makes it clear that
there are two occurrences of the statement with different qualifiers,
but how does the wikidata data model allows me to distinguish between
these two occurrences?

How do I know that:

  P580(start date)  "March 4 1885"

only applies to the first occurrence of the statement, while:

  P580(start date)  "March 4 1893"

only applies to the second occurrence

Re: [Wikidata-l] Questions about statement qualifiers

[Lydia Pintscher]

Hi Antoine,

On Thu, Oct 31, 2013 at 11:39 AM, Antoine Zimmermann
 wrote:
> Hello,
>
>
> I have a few questions about how statement qualifiers should be used.
>
>
> First, my understanding of qualifiers is that they define statements about
> statements. So, if I have the statement:
>
> Q17(Japan)  P6(head of government)  Q132345(Shinzō Abe)
>
> with the qualifier:
>
>  P39(office held)  Q274948(Prime Minister of Japan)
>
> it means that the statement holds an office, right?
> It seems to me that this is incorrect and that this qualifier should in fact
> be a statement about Shinzō Abe. Can you confirm this?

If I understand the example correctly then yes I agree with you.
Qualifiers are there to qualify a statement. So for example to
indicate how a value was measured or from when to when a given
statement was true.

> Second, concerning temporal qualifiers: what does it mean that the "start"
> or "end" is "no value"?  I can imagine two interpretations:
>
>  1. the statement is true forever (a person is a dead person from the moment
> of their death till the end of the universe)
>  2. (for end date) the statement is still true, we cannot predict when it's
> going to end.
>
> For me, case number 2 should rather be marked as "unknown value" rather than
> "no value". But again, what does "unknown value" means in comparison to
> having no indicated value?

Yes. The difference is that we explicitly state something to not be
there or to be there but we don't know more. Example: We know person X
had children but we don't know who they were.

> Third, what if a statement is temporarily true (say, X held office from T1
> to T2) then becomes false and become true again (like X held same office
> from T3 to T4 with T3 > T2)?  The situation exists for Q35171(Grover
> Cleveland) who has the following statement:
>
> Q35171  P39(position held)  Q11696(President of the United States of
> America)
>
> with qualifiers, and a second occurrence of the same statement with
> different qualifiers. The wikidata user interface makes it clear that there
> are two occurrences of the statement with different qualifiers, but how does
> the wikidata data model allows me to distinguish between these two
> occurrences?
>
> How do I know that:
>
>  P580(start date)  "March 4 1885"
>
> only applies to the first occurrence of the statement, while:
>
>  P580(start date)  "March 4 1893"
>
> only applies to the second occurrence of the statement?
> I could have a heuristic that says if two "start date"s are given, then
> assume that they are the starting points of two disjoint intervales. But can
> I always guarantee this?

You can't but this is the best you can do I fear for the moment.


Cheers
Lydia

-- 
Lydia Pintscher - http://about.me/lydia.pintscher
Product Manager for Wikidata

Wikimedia Deutschland e.V.
Obentrautstr. 72
10963 Berlin
www.wikimedia.de

Wikimedia Deutschland - Gesellschaft zur Förderung Freien Wissens e. V.

Eingetragen im Vereinsregister des Amtsgerichts Berlin-Charlottenburg
unter der Nummer 23855 Nz. Als gemeinnützig anerkannt durch das
Finanzamt für Körperschaften I Berlin, Steuernummer 27/681/51985.

___
Wikidata-l mailing list
Wikidata-l@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikidata-l

Re: [Wikidata-l] Questions about statement qualifiers

[Gerard Meijssen]

Hoi,

The head of state is defined on the state involved.. the head of state is
... the prime minister of Japan.

When Mr Abe is the current prime minister, you add "office held" "Prime
minister of Japan" on his item and you can embelish it with the start date,
the predecessor. For previous prime ministers you do the same but add end
date and successor as well as qualifiers.

When I add qualifiers, I will only add them when I know them. For many
historical people a date of death is known but not the date of birth. As
far as I am concerned that is fine.

When an office holder has two terms in office, he holds that office twice.
Consequently you define the held office twice.
https://www.wikidata.org/wiki/Q439237 is an example of the same office held
five times.
Thanks,
 GerardM


On 31 October 2013 11:39, Antoine Zimmermann wrote:

> Hello,
>
>
> I have a few questions about how statement qualifiers should be used.
>
>
> First, my understanding of qualifiers is that they define statements about
> statements. So, if I have the statement:
>
> Q17(Japan)  P6(head of government)  Q132345(Shinzō Abe)
>
> with the qualifier:
>
>  P39(office held)  Q274948(Prime Minister of Japan)
>
> it means that the statement holds an office, right?
> It seems to me that this is incorrect and that this qualifier should in
> fact be a statement about Shinzō Abe. Can you confirm this?
>
>
>
> Second, concerning temporal qualifiers: what does it mean that the "start"
> or "end" is "no value"?  I can imagine two interpretations:
>
>  1. the statement is true forever (a person is a dead person from the
> moment of their death till the end of the universe)
>  2. (for end date) the statement is still true, we cannot predict when
> it's going to end.
>
> For me, case number 2 should rather be marked as "unknown value" rather
> than "no value". But again, what does "unknown value" means in comparison
> to having no indicated value?
>
>
>
> Third, what if a statement is temporarily true (say, X held office from T1
> to T2) then becomes false and become true again (like X held same office
> from T3 to T4 with T3 > T2)?  The situation exists for Q35171(Grover
> Cleveland) who has the following statement:
>
> Q35171  P39(position held)  Q11696(President of the United States of
> America)
>
> with qualifiers, and a second occurrence of the same statement with
> different qualifiers. The wikidata user interface makes it clear that there
> are two occurrences of the statement with different qualifiers, but how
> does the wikidata data model allows me to distinguish between these two
> occurrences?
>
> How do I know that:
>
>  P580(start date)  "March 4 1885"
>
> only applies to the first occurrence of the statement, while:
>
>  P580(start date)  "March 4 1893"
>
> only applies to the second occurrence of the statement?
> I could have a heuristic that says if two "start date"s are given, then
> assume that they are the starting points of two disjoint intervales. But
> can I always guarantee this?
>
>
> Best,
> AZ
>
> --
> Antoine Zimmermann
> ISCOD / LSTI - Institut Henri Fayol
> École Nationale Supérieure des Mines de Saint-Étienne
> 158 cours Fauriel
> 42023 Saint-Étienne Cedex 2
> France
> Tél:+33(0)4 77 42 66 03
> Fax:+33(0)4 77 42 66 66
> http://zimmer.**aprilfoolsreview.com/
>
> __**_
> Wikidata-l mailing list
> Wikidata-l@lists.wikimedia.org
> https://lists.wikimedia.org/**mailman/listinfo/wikidata-l
>
___
Wikidata-l mailing list
Wikidata-l@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikidata-l

[Wikidata-l] Questions about statement qualifiers

[Antoine Zimmermann]

Hello,


I have a few questions about how statement qualifiers should be used.


First, my understanding of qualifiers is that they define statements 
about statements. So, if I have the statement:


Q17(Japan)  P6(head of government)  Q132345(Shinzō Abe)

with the qualifier:

 P39(office held)  Q274948(Prime Minister of Japan)

it means that the statement holds an office, right?
It seems to me that this is incorrect and that this qualifier should in 
fact be a statement about Shinzō Abe. Can you confirm this?




Second, concerning temporal qualifiers: what does it mean that the 
"start" or "end" is "no value"?  I can imagine two interpretations:


 1. the statement is true forever (a person is a dead person from the 
moment of their death till the end of the universe)
 2. (for end date) the statement is still true, we cannot predict when 
it's going to end.


For me, case number 2 should rather be marked as "unknown value" rather 
than "no value". But again, what does "unknown value" means in 
comparison to having no indicated value?




Third, what if a statement is temporarily true (say, X held office from 
T1 to T2) then becomes false and become true again (like X held same 
office from T3 to T4 with T3 > T2)?  The situation exists for 
Q35171(Grover Cleveland) who has the following statement:


Q35171  P39(position held)  Q11696(President of the United States of 
America)


with qualifiers, and a second occurrence of the same statement with 
different qualifiers. The wikidata user interface makes it clear that 
there are two occurrences of the statement with different qualifiers, 
but how does the wikidata data model allows me to distinguish between 
these two occurrences?


How do I know that:

 P580(start date)  "March 4 1885"

only applies to the first occurrence of the statement, while:

 P580(start date)  "March 4 1893"

only applies to the second occurrence of the statement?
I could have a heuristic that says if two "start date"s are given, then 
assume that they are the starting points of two disjoint intervales. But 
can I always guarantee this?



Best,
AZ

--
Antoine Zimmermann
ISCOD / LSTI - Institut Henri Fayol
École Nationale Supérieure des Mines de Saint-Étienne
158 cours Fauriel
42023 Saint-Étienne Cedex 2
France
Tél:+33(0)4 77 42 66 03
Fax:+33(0)4 77 42 66 66
http://zimmer.aprilfoolsreview.com/

___
Wikidata-l mailing list
Wikidata-l@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikidata-l