Extremely low frequency tag issues between HDT and GSD - possible bug?

[AngledLuffa]

There are a couple tags of extremely low frequency in HDT: VAIMP and VMPP. In general the words with those tags, such as "sein" tagged VAIMP, are tagged in other places with a more common tag, such as "sein" being tagged VAFIN most of the time. Are these unusual tags correct or is there an issue somewhere causing them to crop up?

There are also some "unknown" lemmas:

# sent_id = hdt-s134888
# text = Die SPD-Fraktion habe den Entwurf abgelehnt , weil damit eher das Gegenteil des Gewollten erreicht würde , sagte die Abgeordnete Waltraud Schmidt-Sibeth .
14      Gewollten       unknown NOUN    NN      Person=3        12      nmod    _       _

There is also PIDAT, which never occurs in GSD but occurs relatively frequently in HDT. Perhaps this is a tagging issue in GSD? It seems to occur on words where the UPOS takes on different values depending on context in HDT, but is always the same in GSD.

Lastly, PAV in GSD seems to be the same as PROAV in HDT. If there's any way to unify those, that would be great. Easy enough to do on the consumer side, though.

[akoehn]

Hi John, thanks for raising this! automating and documenting the conversion pipeline is on my TODO list but I don’T know when I will find time for that. I will have a look at it then. When converting the HDT to UD we found that the GSD annotation is often not a good source for annotation guidelines as it has been originally annotated in pre-UD times and then moved forward, resulting in (in our POV) incorrect annotations.