UMR approaches annotates pronominal reference and person/number marking through two attributes, :ref-person
for grammatical person information, and :ref-number
for grammatical number marking. These two attributes can apply to any entity concept - but in the available data annotation, these attributes appear only sporadically (if an explicit nominal is overtly marked for plural):
- [en] Bill saw rare birds.plural today. ... the proper noun Bill is not marked wrt number - it seems that singular value, 3rd person is viewed as a default?
In particular, these attributes are used for implicit arguments and for arguments expressed only through verbal cross-referencing - then the NE category person or thing are used for these arguments (together with :ref-person
and :ref-number
, unless this info is available through the coreference relation annotated on the document level (as t/ thing in the following example)).
- [Sanapaná] m-e-hl-t-om-o=hlta (example 3-3-5 (2))
m-e-hl-t-om-o=hlta
NEG-2/3M.IRR-DSTR-eat-TI-IPFV=PHOD
'They did not eat it.'
(e/ entoma-00 'eat'
:actor (p/ person
:ref-person 3rd
:ref-number Plural)
:undergoer (t/ thing)
:aspect Performance
:modstr FullNeg)
The :ref-person
seems to be relevant only for person or thing standing for implicit arguments/arguments expressed through verbal cross-referencing in the released data (!!:refer-person
there!!).
It appears only in those cases where the number cannot be restored via coreference.
The default level annotation (standard 1st, 2nd, and 3rd person values) is appropriate for Czech (and this value can be transferred from the PDT grammateme person
).
The default level for :ref-number
comprises the distinction between singular and non-singular, the latter covering paucal ('a few', a small inexactly numbered group of referents) and plural ('many').
For English, the released data use the :refer-number
attribute (instead of :ref-number
, as in the Guidelines). This attribute seems to be fully annotated with person or thing (used for implicit arguments/arguments expressed through verbal cross-referencing, see above) but only in those cases where the number cannot be restored via coreference (document level annotation). ML: not sure about this - it seems that placeholders like "thing" often do not have this attribute annotated (even if there is no info on coreference there, like in snt2 Estonsko: with ARG3
(thing) for the publication-91 abstract predicate!ú.
In English data, the :refer-number
attribute is used only to mark for non-default value(s), i.e. other than singular number for other entity concepts (esp. unmarked nouns, NE, etc.).
Contrary, for Czech, we have decided to enter it explicitly even for the singular values with all nouns and (personal) pronouns. We apply this also with abstract entities and with named entities (their abstract concepts, not with names).
But: What about, e.g., a NE denoting 1 entity with a plural :name (as in Zbabělci "Cowards" (the book title").
====
As far as I can see, English data are covered be the singular and plural values. The same seems appropriate for Czech where the PDT grammateme number
can be used. This grammateme correctly covers also:
- pluralia tantum (jedny dveře.sg / dvoje dveře.pl)
--> we decided to stick (at least tentatively) to the grammatical number (derived from morphology ... based on the UMR guidelines, Part 3-3-5. Ref: "[the attribute]:refer-number
for grammatical number marking") - polite forms of the 2nd person ("vykání", Vy.sg jste se nepřihlásil?)
- indefinite pronouns (Řekněte, kdo.sg přišel / kdo.pl přišli)
- "container" numerals (Přišlo sto.sg studentů. / Přišlo dvě stě.pl studentů.; probably not relevant for UMR??)
PDT grammateme typgroup
... probably not relevant for UMR now
Wikipedia exemplifies paucal number (among others) on Russian: "In Russian, the genitive singular is also applied to two, three or four items (2, 3, 4 ка́мня – stones, gen. sg.; but 5...20 камне́й – stones, gen. pl.), making it effectively paucal[citation needed] (cf. э́тот ка́мень – this stone, nom. sg.; э́ти ка́мни – these stones, nom. pl.)."