Because prepositions are so important in English, we have a well-established practice of distinguishing ordinary prepositional nmod and obl from other kinds via subtyping (nmod:poss, etc.).
In particular, nmod:tmod/obl:tmod have been used for non-prepositional temporal adjunct nominals like
- It will happen Friday. (
obl:tmod) The party Friday was widely attended. (nmod:tmod)
in contrast to
- It will happen on Friday. (
obl) The party on Friday was widely attended. (nmod)
tmod is part of the legacy of Stanford Dependencies. In light of current UD theory, it is an anomaly where the subtype reflects a semantic but not syntactic distinction (#893). Moreover, it is potentially confusing that only some temporal obliques (the prepositionless ones) receive the subtype.
Meanwhile, nmod:npmod/obl:npmod are used for OTHER non-prepositional adjunct nominals (in special constructions like "5 dollars a share" and "Shares eased a fraction). The term "npmod" (derived from the npadvmod relation in Stanford Dependencies) has been a source of confusion and invokes a concept of NP that is not part of UD theory.
A discussion amongst the core group concluded that a subtype named :unmarked would be a less confusing way to implement the adpositional vs. non-adpositional distinction, for languages that choose to do so.
@amir-zeldes and I plan to implement this for our English corpora, by simply renaming both :tmod and :npmod to :unmarked. Perhaps English-Atis (@aslikuzgun), English-ESLSpok (@kristopherkyle), English-{LinES, Pronouns, PUD} (@AngledLuffa), English-ParTUT (@msang) would like to do so as well for consistency.
Because prepositions are so important in English, we have a well-established practice of distinguishing ordinary prepositional
nmodandoblfrom other kinds via subtyping (nmod:poss, etc.).In particular,
nmod:tmod/obl:tmodhave been used for non-prepositional temporal adjunct nominals likeobl:tmod) The party Friday was widely attended. (nmod:tmod)in contrast to
obl) The party on Friday was widely attended. (nmod)tmodis part of the legacy of Stanford Dependencies. In light of current UD theory, it is an anomaly where the subtype reflects a semantic but not syntactic distinction (#893). Moreover, it is potentially confusing that only some temporal obliques (the prepositionless ones) receive the subtype.Meanwhile,
nmod:npmod/obl:npmodare used for OTHER non-prepositional adjunct nominals (in special constructions like "5 dollars a share" and "Shares eased a fraction). The term "npmod" (derived from thenpadvmodrelation in Stanford Dependencies) has been a source of confusion and invokes a concept of NP that is not part of UD theory.A discussion amongst the core group concluded that a subtype named
:unmarkedwould be a less confusing way to implement the adpositional vs. non-adpositional distinction, for languages that choose to do so.@amir-zeldes and I plan to implement this for our English corpora, by simply renaming both
:tmodand:npmodto:unmarked. Perhaps English-Atis (@aslikuzgun), English-ESLSpok (@kristopherkyle), English-{LinES, Pronouns, PUD} (@AngledLuffa), English-ParTUT (@msang) would like to do so as well for consistency.