Open TomazErjavec opened 3 years ago
Dear Tomaz, You had the right decision. I will look into it tomorrow. My question is, would you like to have the exact affiliations? I can try to see whether it is feasible to find them. Best for now, mc
There are too many speakers (814!) for you to check this, and it is a difficult thing to check. So, pls. don't bother. You should rather enjoy the sun it it will still be shining!
Excellent, then!! mc
Reopening, as we didn't solve the problem, just decided that we will not do it. But maybe somebody someday will, so let's leave it open.
Ok, I see the rationale. Let's leave it open in case we want to modify in the future. Best mc
Is this error the same as the previous one @matyaskopp?
This is related to #19, but more general. We don't have the correct timespans of affiliations with political parties/parliamentary groups. I believe this information can be reachable on Parliament's website, but it is not present in CD format because the metadata in CD format is related to the single date of the meeting (it does not describe the period).
This is an issue that won't be closed soon, if ever. Namely, the person (speaker) affiliations (in particular, being an MP or being a member of a political party) are in ParlaMint marked with dates, i.e. from when to when a person was an MP or was a member of political party, cf. e.g. https://github.com/clarin-eric/ParlaMint/blob/a1110008eae5bc837d111bf46aa405671948fd13/ParlaMint-SI/ParlaMint-SI.xml#L702-L713.
If, during the time that the ParlaMint corpus covers, the person was always and MP or always a member of a party, then the dates are not really necessary, but if they started or stopped being an MP, or changed their party affiliation, they are necessary because we want to be able to say for each speech if the speaker was then an MP or a member of a particlar party.
For the Spanish case I had to hunt for dates when people are MPs or member of parties based on their speaker information associated with each speech. Which means they get as the dates of their affiliations from the first time they spoke to the last time they spoke in the corpus, e.g.: https://github.com/calzada/PARLAMINT-ES-MC/blob/d407704c76a178e10278b973f9e93e4a06786727/ParlaMint/ParlaMint-ES.xml#L689-L694 (But note that I get rid of the dates for party affiliations when a person is a member of only one party.)
Now, this will give the correct values (if I don't have bugs in the program!) of their affiliations for the ParlaMint-ES corpus, but is of course wrong as regards their true dates of affiliation, because they were almost cetainly MPs or party members before they spoke first in the corpus, and after they spoke last.
So, this is just to comment this problem, but I don't see that I can do anything to really solve it - ok, I have some ideas how I could make it a bit better, but it would mean too much programming for too little effect, and the general problem still wouldn't go away..