Feature-decomposition exponence and the Subset Principle #
The realizational engine for privative feature-decomposition accounts of
syncretism ([Cah09], [Bob12]): a cell space is decomposed into
feature sets D : Cell → Finset K, and a rule applies where its feature
specification is contained in the cell's — the Subset Principle. Competition
is [Hal97]'s feature counting: the applicable rule of greatest
specification wins, as the shared core's selectBy on feats.card.
Where the linear containment engine (Exponence/Containment/) orders spans by a
single threshold, this engine orders arbitrary feature sets by inclusion, so
card-maximization is genuine specificity only when the applicable
specifications happen to be nested; the general soundness lives in the
IsWinner strict-competition predicate rather than the core's linear
selectBy_isElsewhereWinner. The linear engine embeds as the chain
decomposition chainDecomp i = Finset.Iic i (Exponence/Containment/Selection.lean).
Main declarations #
Rule— a feature specification plus an exponent;Appliesis Subset containmentpattern— the surface exponent at each cell,selectByonfeats.cardIsWinner— strict feature-count competitionnoABA— the order-theoretic *ABA exclusion for a cell triple whose decomposition nests the middle cell between the outer two
A rule of exponence over a feature decomposition D: a feature
specification and an exponent. Applicability is the Subset Principle — the
rule's features are a subset of the cell's ([Hal97]).
- feats : Finset K
The rule's feature specification.
- exponent : F
The exponent inserted.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- Morphology.Decomposition.instRuleRule = { exponent := Morphology.Decomposition.Rule.exponent, Applies := fun (r : Morphology.Decomposition.Rule Cell D F) (c : Cell) => r.feats ⊆ D c }
Equations
- Morphology.Decomposition.instDecidableRelRuleApplies r c = r.feats.instDecidableLE (D c)
The specificity preorder: applicability-set inclusion, so the engine
participates in the shared core's Elsewhere selection theory
(Exponence/Select.lean). Unlike the linear containment order this is only a
preorder — incomparable specifications (No Case Containment) are the point.
Specificity unfolds to applicability-set inclusion: r ≤ s iff r applies
wherever s does.
The surface pattern of a vocabulary: at each cell, the exponent of the most
highly specified applicable rule — the Elsewhere Principle as selectBy on
feature cardinality ([Hal97]'s counting formulation).
Equations
- Morphology.Decomposition.pattern v c = Option.map Morphology.Decomposition.Rule.exponent (Morphology.Exponence.selectBy (fun (r : Morphology.Decomposition.Rule Cell D F) => r.feats.card) v c)
Instances For
*ABA, order-theoretically: whenever the middle cell y nests between
the outer cells — D y ⊆ D z (appliers at y persist to z) and
D x ∩ D z ⊆ D y (appliers at both outer cells reach y) — a rule winning both
x and z also wins y, so no exponent can interrupt an A_A pattern with a
distinct middle B. The hypotheses are the shared profile of Strong and Weak Case
Containment; No Case Containment violates the first.