Documentation

Linglib.Syntax.HPSG.Categories

HPSG categories: grounding in Universal Dependencies and compatibility #

[Sag10]

The HPSG syntactic categories are the cat sub-hierarchy of the RSRL construct signature (Syntax/HPSG/Construction): cat > {verbal, nonverbal}, verbal > {verb, comp}, nonverbal > {nominal, adj}, nominal > {noun, prep}, with a genuine subsumption order. Empirical data is tagged with Universal Dependencies POS tags (UD.UPOS); udToCat is the grounding bridge, mapping each UD tag to its HPSG cat sort so UD-tagged data enters the category hierarchy by conversion (the project's Layered-Grounding discipline) rather than UD.UPOS being used as the category type.

udToCat is where the nominal conflation becomes principled: NOUN/PRON/PROPN all map to noun, so categoriesMatch treats them alike via the cat hierarchy's ≤ (noun ≤ nominal, etc.) without any hand-coded isNom test. Tags with no HPSG cat correspondent (ADV, DET, NUM, …) map to none, and categoriesMatch falls back to UD-tag equality for them.

source
def HPSG.udToCat :
UD.UPOS → Option Construction.Srt

The HPSG cat sort a Universal Dependencies POS tag grounds to ([Sag10]'s category hierarchy). NOUN/PRON/PROPN → noun (NP), ADP → prep (PP), VERB/AUX → verb, ADJ → adj, SCONJ → comp (complementizer); tags with no cat correspondent are none.

Equations
Instances For
    source
    def HPSG.categoriesMatch (c1 c2 : UD.UPOS) :
    Bool

    Are two categories compatible? Grounded in the HPSG cat hierarchy: each UD tag maps to its cat sort (udToCat) and the two must be comparable in the subsumption order — so all nominal categories (NOUN/PRON/PROPN → noun) match each other without a hand-coded test, and a sort compares with its supersorts. For tags with no cat correspondent it falls back to UD-tag equality.

    Equations
    Instances For