Open olesar opened 7 years ago
ru-syntagrus: 5 более более ADV Degree=Cmp 7 nummod:gov 7:nummod:gov 6 пяти пять NUM Case=Gen 7 nummod 7:nummod 7 лет год NOUN _ Animacy=Inan|Case=Gen|Gender=Masc|Number=Plur 4 obl 4:obl SpaceAfter=No
1 Более более ADV Degree=Cmp 4 nsubj 4:nsubj 2 двух два NUM Case=Gen 3 nummod 3:nummod 3 месяцев месяц NOUN Animacy=Inan|Case=Gen|Gender=Masc|Number=Plur 1 nmod 1:nmod 4 прошло проходить VERB Aspect=Perf|Gender=Neut|Mood=Ind|Number=Sing|Tense=Past|VerbForm=Fin|Voice=Act 0 root 0:root
21 больше много NUM 23 nummod:gov 23:nummod:gov 22 300 300 NUM 23 nummod 23:nummod 23 заявок заявка NOUN Animacy=Inan|Case=Gen|Gender=Fem|Number=Plur 20 obl 20:obl (NB different pos)
более/больше/менее/меньше should be linked to the numeral head, cf. террористов там было не более двух.
I think that the ordinals are compounds:
compound(второго, сорок) amod(года, второго)
Related to numerals is https://github.com/UniversalDependencies/docs/issues/455.
nmod (dep?) depending on ADJ or ADV --> obl
If the ADJ or ADV is a head of copula construction then you are right: such ADJ|ADV should not have nmod
children, but obl
.
In the remaining cases, we should be careful: the ADJ could be a head of a noun phrase with elided noun and then nmod
child is correct.
BTW: This is exactly the case when the nmod
vs. obl
distinction is needed because it cannot be reconstructed fully automatically (at least not easily).
acl with participles (single participles vs. prtcp group), advcl vs. acl. Need attention.
discourse/parataxis is tagged differently in two treebanks ==> Cross-Check Task, scheduled March 2018 ==> Olga makes two lists
vocative: check parataxis & NOUN & Animacy=Anim in ru-SynTagRus