CRDC: Conjunct Referential Dependency Constraint #
The Conjunct Referential Dependency Constraint of [OL23], p. 651 (verbatim):
A referentially dependent conjunct valent can be co-valued with a full co-valent, but a referentially dependent full valent can hardly be co-valued with a conjunct co-valent.
The CRDC governs only configurations in which one of the relevant
positions sits inside a coordinate structure; non-coordinate binding
falls under Conditions A/B/C, on which the paper is explicit that the
CRDC is silent (p. 651). Marginality is constitutive of its empirical
content: the paper's crowdsourced acceptability table (p. 630 fn. 3)
maps mean scores to markers — ? (1.65–2.29), ?? (2.30–2.94), *
(2.95–4.00) — and the CRDC's prediction is ??,
Judgment.questionable in the project's enum.
Each theorem builds the tree of a stimulus from
Data.Examples.OsborneLi2023 and compares crdcPredictedJudgment
with the row's recorded judgment: equality where the CRDC is the
operative principle, and a recorded divergence for ex9b, whose
sentence-level marginality is Condition B's contribution.
Implementation notes #
- "Valent" is operationalised as a direct UD valency-relation dependent
of the predicate (
UD.DepRel.isValencyArg) — a deliberate simplification of the paper's catena-based notion (§4); the example set does not exercise the difference. - UD's basic-tree convention makes the first conjunct head the
coordinate structure, with remaining conjuncts attached via
.conj; the conjunct helpers are two-liners overGraph.children. - Binding theories elsewhere in linglib (
Studies/Chomsky1981.lean,Syntax/HPSG/Coreference.lean) make categorical predictions on non-coordinate stimuli; the CRDC contributes a graded prediction on coordinate ones. A head-to-head comparison needs coordination-aware binding parsers andJudgment-valued output on their side.
TODO #
- Cover §6's counterexamples (e.g. the vote-predicate identity split, ex (55a) in the data) with a third-party-referent treatment.
Predicate-valent type #
The conjuncts of the coordinate structure headed at c: the head
(the first conjunct, per UD) plus its .conj dependents.
Equations
- OsborneLi2023.allConjuncts g c = insert c ({w ∈ g.children c | g.label c w = some UD.DepRel.conj})
Instances For
Position c heads a coordinate structure: it has a conj dependent.
Equations
- OsborneLi2023.HasConjuncts g c = ∃ w ∈ g.children c, g.label c w = some UD.DepRel.conj
Instances For
Equations
- OsborneLi2023.instDecidableHasConjuncts g c = Finset.decidableExistsAndFinset
Word valentIdx is a conjunct valent of predicate predIdx: a
conjunct of a coordinate structure that fills a valency role of
predIdx.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- OsborneLi2023.instDecidableIsConjunctValent g predIdx valentIdx = Finset.decidableExistsAndFinset
Word valentIdx is a full valent of predIdx: a valent that is
not a conjunct valent — the paper's definition (p. 651): "a valent
of a given predicate is a full valent thereof if it is complete,
that is, it is not a conjunct valent."
Equations
- OsborneLi2023.IsFullValent g predIdx valentIdx = ((∃ r ∈ g.label predIdx valentIdx, r.isValencyArg = true) ∧ ¬OsborneLi2023.IsConjunctValent g predIdx valentIdx)
Instances For
Equations
- OsborneLi2023.instDecidableIsFullValent g predIdx valentIdx = OsborneLi2023.instDecidableIsFullValent._aux_1 g predIdx valentIdx
CRDC prediction #
The CRDC's predicted judgment for co-valuing anaphor anaIdx with
antecedent anteIdx under predicate predIdx: .questionable
exactly when the anaphor is a full valent and the antecedent a
conjunct valent of the same predicate; .acceptable otherwise —
the CRDC is silent, and other binding principles may still apply.
Equations
- One or more equations did not get rendered due to their size.
Instances For
CRDC predictions against the data rows #
Ex (2a): "Max and Lucie talked about him."
Max(0) and(1) Lucie(2) talked(3) about(4) him(5).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Full-valent him over conjunct-valent Max: the CRDC fires and
matches the observed ??.
Ex (9a): "Max talked about himself." — non-coordinate baseline.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Ex (9b): "Max talked about him." — non-coordinate Condition B context. The CRDC is silent; the row's marginality is Condition B's contribution, so prediction and row diverge by design.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Ex (24a): "John talked about himself and his mother." — coordinate object;
himself heads the coordination, so it is a conjunct valent and
John a full valent: the CRDC's permitted direction.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Ex (5a): "Both John and Mary love him." — coordinate subject with paired
coordinator; pronoun in object position. The CRDC fires and matches
the observed ??.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Ex (28d): "John expected Mary and him to be able to leave soon." —
coordinate object under raising-to-object: him is a conjunct
valent, so the CRDC is silent on co-valuing it with John.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Directionality #
The CRDC is asymmetric: on ex2a's tree, only the full-anaphor-of- conjunct-antecedent direction fires; swapping anaphor and antecedent leaves the CRDC silent.