Dutch test set

[desidoicheva]

Hi all, in case you have not noticed yet and you have some problems with the Dutch test set, there is a very tiny formatting difference that you might want to consider - opposite to everything we have had by now, each of the lines does not start with a number but with 3 spaces and then a number.

Good luck!
Desi

[desidoicheva]

I just found nother problem with the Dutch test set ... It concerns the following sentence in document 43:

1 ' _ ' _ LET _ _ _ 0 _ ROOT _ _ _ _ _
2 Waarom _ Waarom _ BW _ _ _ 0 _ ROOT _ _ _ _ _
3 bewijst _ bewijzen _ WW _ mood=finiteform|tense=present|number=with-t _ 2 _ body _ _ _ _ _
4 ze _ ze _ VNW _ type=personal|pdtype=pron|case=standard|person=3|number=singular|gender=fem _ 3 _ su _ _ _ _ (133)
4 niet _ niet _ BW _ _ _ 3 _ mod _ _ _ _ _
5 dat _ dat _ VG _ type=subord _ 3 _ vc _ _ _ _ _
6 de _ de _ LID _ type=definite|case=standard _ 8 _ det _ _ _ _ _
7 feiten _ feit _ N _ type=common|num=plural|degree=basis _ 9 _ cnj _ _ _ _ _
8 en _ en _ VG _ type=coord _ 15 _ su _ _ _ _ _
9 uitspraken _ uitspraak _ N _ type=common|num=plural|degree=basis _ 9 _ cnj _ _ _ _ _
10 van _ van _ VZ _ position=initial _ 8 _ mod _ _ _ _ _
11 ons _ ons _ VNW _ type=possessive|pdtype=det|case=standard|person=1|number=plural|position=prenominal|inflection=without _ 13 _ det _ _ _ _ (134
12 programma _ programma _ N _ type=common|num=singular|degree=basis|gender=neutral|case=standard _ 11 _ obj1 _ _ _ _ 134)
13 fout _ fout _ ADJ _ position=free|degree=basis|inflection=no _ 15 _ predc _ _ _ _ _
14 zijn _ zijn _ WW _ mood=infinitive|position=free _ 6 _ body _ _ _ _ _
15 ? _ ? _ LET _ _ _ 15 _ punct _ _ _ _ _
16 ' _ ' _ LET _ _ _ 15 _ punct _ _ _ _ _


The problems are:
1. There are two words with the same number - #4
2. Words 9 and 15 have recursive dependencies.

Best
Desi

[emili]

Ok, I will try to contact with the dutch dataset responsibles. Anyway, we are aware that the time to solve the test is limitated so try to solve as many documents as possible and leave the problematics apart. If the problematic documents are not fixed soon, we will ignore them in the final scores.

Thank you for the advice.

[veronique]

Hi Desi,

there is indeed a double ID in that sentence. If you ignore the second 4 and
number all words up to 17, all dependencies are OK.

Best,
Veronique

[desidoicheva]

Hi again,

just for info - I found out that the test set has a document (#21) that does not have any coreference markings. This again leads to problem with the scorer. How was this solved for the final scoring actually?

Best
Desi