search

IAHLT / UD_Hebrew

Hebrew Universal Dependencies Treebank
Other
2 stars 2 forks source link

ADV in nismax position #35

Closed Hilla-Merhav closed 3 years ago

Hilla-Merhav commented 3 years ago

@amir-zeldes

How do you think we should analyze expressions like למחרת החקירה?

Usually as far as I know compound is deprel only NOUNs (and sometimes maybe ADJs) govern through, but למחרת is an ADV.

The HTB solution – in its one and only occurrence – is fixeding it. But since every NOUN that refers to any event can occur in this environment, this solution doesn't look applicable to me.

Also, I think this question connects to the לבד+ו issue (#32) – this is a general question about the connection between a governor ADV and a dependent NOUN/PRON.

amir-zeldes commented 3 years ago

TBH I think it should be segmented, so you get back to le+maxarat, and then everything is OK. Can we treat it as the construct state form of maxar in this case? We can still decide separately if we like "lemaxarat" when it appears by itself as an ADV (this is like how "ha-yom" is one token for "today" and two for "the day")

Hilla-Merhav commented 3 years ago

@amir-zeldes Sounds good. In Even Shoshan I see that מחר is both adverb and masculine noun, but its nismax form is מְחַר־, and מחרת is a separate lexical item – also both adverb and noun, but a feminine noun, so I suggest the lemma of מחרת would be מחרת and we assign to it the feature Gender=Fem, is it OK?

amir-zeldes commented 3 years ago

OK!