NP-POS

From Icelandic Parsed Historical Corpus (IcePaHC)
Jump to: navigation, search

NP-POS is the phrase-level tag for genitive NPs which occur within other NPs. It does *not* occur at the IP level of structure or at the PP level; genitive objects of P are bare NPs. Note that genitive NPs inside an ADJP are labeled NP-POS if there is a possessive relationship there, e.g. "(ADJP fegurst (NP-POS fjallanna))". There is no -POS extension on NPs where there is no possessive relationship, e.g. "(ADJP fagurt (NP álits))"..

We also parse all inherently possessive pronouns as projecting an NP-POS, even if they are not in the genitive form of the possessive pronoun. E.g. "minni" projects an NP-POS in "bókinni minni", even though it is the dative form of the inherently possessive pronoun.

(Note: the Old Icelandic form "voru" is one of these: it's the dative form of an inherently possessive pronoun the normal dative of the 1pl pronoun is "oss" in OIce., as in "voru bæn". If it were "bæn manninum" or "bæn oss", then it would be a dative of possession and project an NP-COM). See NP-COM for details about non-genitive possessors which are not inherently possessive pronouns.

All genitive nouns which occur within another NP, e.g. possessors, project an "NP-POS", whether or not they are single words. This is consistent with the standard of the York Corpus of Old English Prose (YCOE) and the PPCME2 with respect to postnominal possessors, though those corpora differ from ours w.r.t. prenominal possessors. In the Icelandic corpus, prenominal possessors also project their own "NP-POS", consistent with the conventions in Caitlin Light's corpus of Early New High German.

However, we do *not* project an NP-POS when an NP contains an N and a D (and possibly ADJ) and these are genitive because they have all been assigned genitive from an external assigner (i.e. a verb or preposition). In this case, there is no NP-internal relationship between the N and D (or N, D, and ADJ); they are both marked genitive simply because they must agree in case. Thus, the following is a correct parse:

(PP (P til-til)
    (NP (Q-G hverrar-hver) (N-G hátíðar-hátíð))

THE ABOVE PHRASE SHOULD *NOT* BE PARSED AS:

(PP (P til-til)
    (NP (NP-POS (Q-G hverrar-hver)) (N-G hátíðar-hátíð))
  • NOR* SHOULD IT BE PARSED AS:
(PP (P til-til)
    (NP-POS (NP-POS (Q-G hverrar-hver)) (N-G hátíðar-hátíð))

...because there is no possession relationship between the Q and N, nor is there any adnominal relationship other than the usual determiner-noun relationship.


Since the distinction between genitives of possession and other genitive constructions is not always clear in Icelandic, we do not try to distinguish the two: all genitive NPs occurring within another NP are labelled NP-POS, whether or not they are truly possessive. For NP complements of N which have other case-marking, see NP-COM.

So, for clause level genitive NPs:

(NP-OBJ (N-...G..) )

And for genitives within another NP:

(NP-POS (N-...G..) )

In both cases, genitive Case is indicated on the word-level dash-tag.

No IP-level NP-POS

At the IP level, NPs are tagged for their grammatical relation, e.g. NP-SBJ, NP-OB1, not for their Case, so "NP-GEN" does not occur in the corpus.

Note applying to Determiners, Adjectives, Nominals: we have adopted, across the board, a policy of tagging Case on the word-level, but not on the phrase level (grammatical relations are tagged on the phrase level). So if you are searching for Subject, Object, etc., search for the phrase labels, but if you search for Case, you'll have to indicate the appropriate word-level tag.

However, there are some cases where NP-POS projects an IP-level NP without any head noun (there may be an understood silent noun). These are usually cases of NP-PRD, and in these cases the NP-POS usually can be translated as an "of"-PP in English, as in the following example meaning "we are of this world":

	                  (IP-SUB (NP-SBJ (PRO-N vér-ég))
					(BEPI erum-vera)
					(NP-PRD (NP-POS (D-G þessa-þessi) (N-G heims-heimur)))))))

The above sentence could also be paraphrased, "we are (people) of this world".

No PP-level NP-POS

See complements of PP.

Recursive NP-POS

Genitive/possessive structures can also be recursive, of course. According to the general policy, we try to minimize the amount of structure that individual words project where possible, even though genitive nouns within another NP always project at least one NP-POS. Thus, in the case of recursive genitive structures, there is a top-level NP-POS, which immediately dominates a genitive noun and another NP-POS (i.e., the embedded NP-POS and genitive noun are sisters), and so on, as in the following example:

	  (NP-VOC (N-N dalur-dalur)
		  (NP-POS (N-G fósturjarðar-fósturjörð)
			  (NP-POS (PRO-G minnar-minn))))


Note that there will be some possessive pronouns which are not genitive, but still project an NP-POS for the sake of parallelism, as in the example below:

( (IP-MAT (CONJ og-og)
	  (NP-SBJ *con*)
	  (HVDI hafði-hafa)
	  (ADVP (ADVS mest-meira))
	  (NP-OB1 (NS-A mök-mök))
	  (PP (P við-við)
	      (NP (NP-POS (PRO-A sína-sinn))
		  (NS-A sveitunga-sveitungi)))
	  (. .-.))
  (ID PILTUR,.37))