search

clarin-eric / ParlaMint

ParlaMint: Comparable Parliamentary Corpora
https://clarin-eric.github.io/ParlaMint/
43 stars 53 forks source link

ParlatMint-AT: Validation WARN: political parties #526

Closed hpreki closed 1 year ago

hpreki commented 1 year ago

There a some warnings left in the ParlaMint-AT validations. I am very tempted to simply ignore them all :-)

But for 2 WARN-types I think I even can bring up some reasons to do so.

@TomazErjavec and @matyaskopp you might remember the longish discussions we had about encoding party-affiliations and how to deal with the non-existing temporal information. I ended up with using the exactly timed member-relation between MP and (party-related) parliamentaryGroup . And I use membership to politicalParty when a person (throughout their political vita) also has a function other than MP: then, and only then, I add a membership to a politicalParty - and these memberships never have any from or to.

This triggers the following warnings:

WARN[08] ParlaMint-AT-listPerson Affiliate from date (1996-01-15T00:00:00) is P3380D before politicalParty.BZÖ organization beginning (2005-04-17T00:00:00) 
WARN[08] ParlaMint-AT-listPerson Affiliate from date (1996-01-15T00:00:00) is P6585D before politicalParty.NEOS organization beginning (2014-01-25T00:00:00)

These 2 parties are younger than the "corpus-start" (1996-01-15) - so obviously you interpret the "non-existing from-to" as from=<CORPUS_START_DATE> to=<CORPUS_END_DATE> and produce a (unjustified) warning.

The other WARN is:

WARN[10] ParlaMint-AT-listOrg politicalParty-role organisation without affiliation: #politicalParty.JETZT
WARN[10] ParlaMint-AT-listOrg politicalParty-role organisation without affiliation: #politicalParty.STRONACH

These 2 parties were small and short-lived - so none of their members made it into a political function outside the parliament. So yes: no person has a direct affiliation to either politicalParty.JETZT or politicalParty.STRONACH . But still I need these entities, because they are needed as "closest relatives" of their respective parliamentaryGroup: 

             <relation name="representing" active="#parliamentaryGroup.JETZT"
                passive="#politicalParty.JETZT"/>
             <relation name="representing" active="#parliamentaryGroup.STRONACH"
                passive="#politicalParty.STRONACH"/>

Summary: unless you protest, I would ignore this WARNings - and maybe you even want to relax / modify your validation process to downgrade these WARN to INFO.

matyaskopp commented 1 year ago

I am very tempted to simply ignore them all :-)

Yes, I agree with you.

This triggers the following warnings:

WARN[08] ParlaMint-AT-listPerson Affiliate from date (1996-01-15T00:00:00) is P3380D before politicalParty.BZÖ organization beginning (2005-04-17T00:00:00) 
WARN[08] ParlaMint-AT-listPerson Affiliate from date (1996-01-15T00:00:00) is P6585D before politicalParty.NEOS organization beginning (2014-01-25T00:00:00)

These 2 parties are younger than the "corpus-start" (1996-01-15) - so obviously you interpret the "non-existing from-to" as from=<CORPUS_START_DATE> to=<CORPUS_END_DATE> and produce a (unjustified) warning.

You are right. In the beginning, there was an agreement that if a date is missing whole corpus period is used, but it can also happen that the date is missing because it is not possible to get it.

Summary: unless you protest, I will ignore these WARNings - and maybe you even want to relax / modify your validation process to downgrade these WARN to INFO.

No protest, but I am unsure if I want to change these WARNings into INFO because no one reads INFO messages. The warnings try to highlight that there can be something odd and the corpus maker should think about it (As you did :-) )

I guess @TomazErjavec has no complaints, so you are free to close.