Vocabularies over containment hierarchies: rules and specificity #
The carrier for the realizational engine of [Bob12]'s
comparative-suppletion generalizations, over an arbitrary n-grade
containment hierarchy. An SpanRule realizes the initial span [0, spans]
of the hierarchy, optionally conditioned on a higher head. One carrier
carries two Exponence.Rule views: the Subset reading (SpanRule,
applicability is threshold containment) is DM Elsewhere insertion; the
Superset reading (SupersetRule, applicability is constituent
containment) is nanosyntax spellout, dual à la OrderDual.
Main declarations #
SpanRule— exponent, exponed span[0, spans], optional conditioning headSpanRule.threshold,AppliesAt,MoreSpecific— Subset applicability and itsmoreSpecific_iff_threshold_lereflectionSpanRule.Matches,SupersetRule— Superset applicability and itsmatches_imp_iff_spans_lereflectionTerminal,Adjacent,Antihomophonous,Grounded,ContextFree— well-formedness conditions on vocabularies
Rules of exponence and derived specificity #
A rule of exponence ([Bob12]) over an n-grade containment
hierarchy. The rule realizes the initial span [0, spans] — the bare root
when spans = 0, a root+heads portmanteau when spans > 0 — and applies
only when its optional conditioning head context is present. Latin
([Bob12] (204)): bon is ⟨bon, 0, none⟩, mel- is
⟨mel, 0, some 1⟩, the portmanteau opt- is ⟨opt, 1, some 2⟩.
- exponent : F
The phonological exponent inserted for the span.
- spans : Fin n
Upper end of the exponed initial span
[0, spans]. - context : Option (Fin n)
Head whose presence conditions the rule, if any.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
The least grade at which the rule is applicable: everything the rule mentions — exponed span and conditioning context — must be contained in the structure.
Instances For
Equations
A rule it is at least as specific as jt when it applies in a
subset of the contexts jt applies in (Pāṇinian specificity).
Equations
- it.MoreSpecific jt = ∀ ⦃g : Fin n⦄, it.AppliesAt g → jt.AppliesAt g
Instances For
Over a linear containment hierarchy, applicability sets are nested upward sets, so derived specificity is threshold comparison — the Elsewhere ordering ([Kip73]) is total.
Well-formedness conditions on vocabularies #
Each generalization below hypothesizes exactly the conditions it
needs; vocabularies violating a condition witness the corresponding
unattested pattern (see the worked examples in
Studies/Bobaljik2012.lean).
Every rule expones the bare root (no portmanteaux).
Equations
- Morphology.Containment.Terminal v = ∀ (it : Morphology.Containment.SpanRule n F), it ∈ v → ↑it.spans = 0
Instances For
Equations
- Morphology.Containment.instDecidableTerminal v = id inferInstance
Conditioning heads are adjacent to the exponed span — [Bob12]'s (tentatively adopted) adjacency condition on contextual allomorphy.
Equations
- Morphology.Containment.Adjacent v = ∀ (it : Morphology.Containment.SpanRule n F), it ∈ v → ∀ (c : Fin n), it.context = some c → ↑c = ↑it.spans + 1
Instances For
Equations
- Morphology.Containment.instDecidableAdjacent v = id inferInstance
Distinct rules carry distinct exponents — [Bob12]'s Antihomophony assumption (44), closing the loophole where a surface-ABA pattern is really an ABC with accidental A ≡ C homophony.
Equations
- Morphology.Containment.Antihomophonous v = ∀ (it : Morphology.Containment.SpanRule n F), it ∈ v → ∀ (jt : Morphology.Containment.SpanRule n F), jt ∈ v → it.exponent = jt.exponent → it = jt
Instances For
Equations
- Morphology.Containment.instDecidableAntihomophonousOfDecidableEq v = id inferInstance
[Bob12]'s markedness condition (202): a context-sensitive rule of exponence involving a node requires a context-free rule involving that node. Under the threshold encoding, this is downward closure of the vocabulary's threshold set.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- Morphology.Containment.instDecidableGrounded v = id inferInstance
The nanosyntax restriction, in the pointer-free idealization of [Cah09]: entries store bare constituents, with no DM-style contextual environment.
Equations
- Morphology.Containment.ContextFree v = ∀ (it : Morphology.Containment.SpanRule n F), it ∈ v → it.context = none
Instances For
Equations
- Morphology.Containment.instDecidableContextFree v = id inferInstance
The Subset reading: DM Elsewhere insertion #
The containment engine implements Morphology.Exponence.Rule
directly: applicability is threshold containment (the upper set
Set.Ici threshold) and derived specificity is threshold comparison.
Equations
- Morphology.Containment.instRuleSpanRuleFin = { exponent := Morphology.Containment.SpanRule.exponent, Applies := fun (it : Morphology.Containment.SpanRule n F) (i : Fin n) => it.threshold ≤ i }
Equations
- Morphology.Containment.instDecidableRelSpanRuleFinApplies it g = it.threshold.decLe g
Subset applicability is threshold containment.
Containment specificity is the shared core's specificity order.
The Superset reading: nanosyntax spellout #
One carrier, a dual view: the entry can spell out grade g when its
stored constituent contains g (the down-set Set.Iic spans), and
smallest-match selection is Pāṇinian specificity under superset
matching. A distinct type synonym so the two readings carry different
Exponence.Rule instances on one carrier, mirroring OrderDual.
The Superset Principle ([Sta09a]): an entry can spell out
grade g when the constituent it stores contains grade g's
structure. Anti-monotone in g, dually to AppliesAt.
Instances For
Minimize Junk derived, dually to
SpanRule.moreSpecific_iff_threshold_le: an entry matches in a
subset of the structures another matches in iff it stores the smaller
constituent — smallest-match selection is Pāṇinian specificity under
superset matching.
Equations
- One or more equations did not get rendered due to their size.
Equations
- Morphology.Containment.instDecidableRelSupersetRuleFinApplies it g = g.decLe it.spans
Read an exponence rule under the Superset reading.
Equations
- it.superset = it
Instances For
Superset applicability is constituent containment.
Minimize Junk is the Superset reading's specificity order: smaller span = more specific.
Subset/Superset duality over context-free vocabularies (the
nanosyntax idealization): DM-style Subset specificity of it over jt
is Superset specificity of jt over it. With contextual restrictions
the Subset order compares thresholds, not spans, and the duality is
only one-directional.