Documentation

Linglib.Theories.Syntax.CCG.CrossSerial

Verb phrase (infinitival)

Equations
Instances For

    Control verb: takes VP, gives VP (e.g., "helpen" = help)

    Equations
    Instances For

      Perception verb: takes VP, gives TV-like (e.g., "zag" = saw)

      Equations
      Instances For

        Infinitival transitive: VP/NP (e.g., "zwemmen" when taking object)

        Equations
        Instances For

          Infinitival verb needing subject (for cross-serial): (S\NP)/NP

          In Dutch verb-raising constructions, the infinitive's subject appears in what looks like an object position. This category allows the embedded subject to be picked up via composition.

          zwemmen : (S\NP)/NP — "swim" needing its subject as argument

          This differs from the simple intransitive VP = S\NP.

          Equations
          Instances For

            Verb-raising control verb: ((S\NP)/NP)/(S\NP)

            In verb-raising constructions with 3+ verbs, each intermediate restructuring verb (helpen, laten) must provide an extra /NP slot for its own raised subject. Standard ControlV = (S\NP)/(S\NP) only passes through the complement's arguments; ControlVR adds an /NP so the verb's subject can be picked up via composition.

            This is what enables the full cross-serial derivation:

            • zwemmen : (S\NP)/NP — base: needs subject
            • helpen : ((S\NP)/NP)/(S\NP) — VR: needs complement + passes NP
            • zag : (S\NP)/(S\NP) — matrix: standard perception verb
            Equations
            Instances For

              Forward Composition² (B²): X/Y (Y/Z)/W → (X/Z)/W

              Allows "reaching through" one extra argument.

              Equations
              Instances For

                Forward Composition³ (B³): X/Y ((Y/Z)/W)/V → ((X/Z)/W)/V

                Allows "reaching through" two extra arguments.

                Equations
                Instances For

                  Extended derivation with generalized composition.

                  Instances For
                    Equations
                    • One or more equations did not get rendered due to their size.
                    Instances For

                      Get category from extended derivation

                      Equations
                      Instances For
                        Equations
                        • One or more equations did not get rendered due to their size.
                        Instances For

                          zwemmen with InfSubj category for verb-raising

                          Equations
                          Instances For

                            helpen with ControlVR category for verb-raising (3+ verb clusters)

                            Equations
                            Instances For

                              laten with ControlVR category for verb-raising

                              Equations
                              Instances For

                                Full cross-serial derivation for 3 verbs #

                                For "Jan Piet Marie zag helpen zwemmen" (Jan saw Piet help Marie swim), the cross-serial bindings are: Jan→zag, Piet→helpen, Marie→zwemmen.

                                The key is the verb-raising category ControlVR = ((S\NP)/NP)/(S\NP): helpen in verb-raising passes through an /NP slot for its raised subject (Piet), in addition to taking its VP complement. This lets B² thread both Piet's and Marie's argument slots through the verb cluster:

                                helpen_vr >B zwemmen_vr : ((S\NP)/NP)/(S\NP) >B (S\NP)/NP = ((S\NP)/NP)/NP
                                zag >B² result          : (S\NP)/(S\NP) >B² ((S\NP)/NP)/NP = ((S\NP)/NP)/NP
                                result + Marie          : ((S\NP)/NP)/NP + NP = (S\NP)/NP       — Marie→zwemmen
                                result + Piet           : (S\NP)/NP + NP = S\NP                 — Piet→helpen
                                Jan + result            : NP + S\NP = S                          — Jan→zag
                                

                                The outermost /NP originates from zwemmen_vr, so Marie binds with zwemmen. The inner /NP originates from helpen_vr, so Piet binds with helpen. The \NP is zag's subject slot, so Jan binds with zag.

                                The 2-NP cross-serial derivation yields category S.

                                The 3-NP cross-serial derivation yields category S. This uses B (forward composition) and B² (generalized composition) to thread all three NP argument slots through the verb cluster.

                                Intermediate: the verb cluster composes into ((S\NP)/NP)/NP, a 3-place predicate needing Jan (\NP), Piet (/NP), Marie (/NP).

                                def CCG.CrossSerial.crossSerialLanguage (n : Nat) :
                                List String

                                The cross-serial language {aⁿbⁿcⁿ | n ≥ 1} is NOT context-free.

                                CCG can generate this via generalized composition, proving CCG is mildly context-sensitive.

                                Equations
                                Instances For