Cooper's type theory with records #
Cooper's theory of types with records has Lean's own type theory as its metatheory: the
judgement a : T is the ambient typing, a type is true when inhabited, a record type is a
structure whose structural subtypes are the structures with more fields, and the
intensionality of types with the same witnesses is the ambient theory's as well. On this
basis the file follows the book's chapters: the contents of proper names, the indefinite
article and the copula, with a is a P witnessed exactly when P(a) is (Ch. 3), and
parametric content (Ch. 4); modal type systems with their restrictive and inclusive notions,
necessity and possibility relative to a background type and a topos in place of an
accessibility relation, and intensionality as the matching of types against an agent's
long-term memory, religious beliefs and desires through points of view (Ch. 6); restricted
properties and their two purifications, the frequentist probability of a witness set and its
estimate from an experience base, and the witness conditions whose witnesses carry what
discourse anaphora picks up (Ch. 7); and parametric contents over a context type of pronoun
labels, on which storage and retrieval derive the scope readings, anaphoric combination
with the sentence boundary and reflexivisation derive Principles B and A, and localisation
derives the weak and strong donkey readings (Ch. 8). A selection of the book's English
examples are the rows of Data/Examples/Cooper2023.json, against which the anaphora sets
the substrate derives for each quantifier are checked.
Implementation notes #
- A property is
E → Typewith Cooper's record[x : Ind]collapsed to its individual, and a ptypep(a)a type of situations, so the common-noun content is the identity and unnamed. Relabelling is absorbed into the function witnessing a subtyping, and compatibility and the topos conditions are read in the ambient possibility. - The indefinite article and no are the particular witness conditions of
Semantics/Quantification/Witness.lean, and the dogs fragment uses its set-based conditions over decidable predicates; modal type systems and Breitholtz's topoi stay here, having no second consumer. - A context type records the labels a content requires of the assignments to stored
quantifiers, pronouns, local pronouns and reflexives; the wh-phrase and gap assignments of
(82) and the incrementation of labels at combination are omitted, labels being chosen
distinct by hand. Context specification
c[𝔰.xᵢ = a]is the one primitive behind retrieval, reflexivisation and cross-sentential resolution. - Restriction and alignment of a property's domain (Ch. 8, (51)–(52)) are further restrictions through which the body is read, so purification applies to them uniformly.
TODO #
- Inclusive possibility is stated as the book prints (2d), an implication under an existential, which any possibility in which the type does not occur satisfies; the intended clause is presumably a conjunction.
- The closure of a content type under the operations (Ch. 8, (21), (89)) and the combination of content types (23)–(24) are not modelled: each reading is derived by composing the operations by hand.
- The number of a pronoun is not modelled, so the contrast between it and they with a universal antecedent, (46b)–(46e), is recorded in the rows but not derived.
- A reflexive inside a verb phrase (The guru revealed Kim to himself) and picture noun phrases are outside the treatment, as in the book.
- Purification is determiner-independent, so both donkey readings are predicted for every determiner; the experimental record of [denic-sudo-2022] on non-monotonic determiners and the question-based selection of [CBH19] are the tests to state.
References #
- R. Cooper, From Perception to Communication (2023)
- J. Barwise, R. Cooper, Generalized Quantifiers and Natural Language (1981)
- [A. Kratzer, What 'must' and 'can' must and can mean (1977)][kratzer-1977]
- [A. Kratzer, The Notional Category of Modality (1981)][kratzer-1981]
- E. Breitholtz, Enthymemes and Topoi in Dialogue (2020)
- [G. Chierchia, Dynamics of Meaning (1995)][chierchia-1995b]
- M. Kanazawa, Weak vs. Strong Readings of Donkey Sentences (1994)
- [A. Ranta, Type-Theoretical Grammar (1994)][ranta-1994]
- R. Montague, The Proper Treatment of Quantification in Ordinary English (1973)
- [M. Denić, Y. Sudo, Donkey Anaphora in Non-Monotonic Environments (2022)][denic-sudo-2022]
- L. Champollion, D. Bumford, R. Henderson, Donkeys under Discussion (2019)
Types, properties and contents (Chs. 1, 3, 4) #
Structural subtyping (§1.4.3.5) #
The type of situations in which the boy hugs the dog, (53b): a subtype of (53a) by having
more fields, the projection being toBoyAndDog.
Instances For
Properties, quantifiers and their contents (§3.4) #
A property (30): the individuals' types of situations.
Equations
- Cooper2023.Ppty E = (E → Type)
Instances For
A quantifier: a function from properties to types, Montague's ⟨⟨e,t⟩,t⟩.
Equations
- Cooper2023.Quant E = (Cooper2023.Ppty E → Type)
Instances For
SemPropName(a) (33): the quantifier applying its property to the individual.
Equations
- Cooper2023.SemPropName a P = P a
Instances For
SemIndefArt (37): a restrictor property to the existential quantifier over it, whose
witness under the particular condition of Ch. 7 (63) is an individual with the restrictor
and the scope.
Equations
- Cooper2023.SemIndefArt restr = Quantification.ParticularWC_Exist restr
Instances For
(55): exist(P, Q) is witnessed iff the property extensions of P and Q overlap.
SemBe (78), Montague's copula: the property of being the quantifier's witness.
Equations
- Cooper2023.SemBe Q x = Q fun (y : E) => PLift (x = y)
Instances For
The universal quantifier as a function from the restrictor's witnesses to the scope's,
the function witness of §7.2.4 after [ranta-1994], which, as Cooper notes at (27), yields no
witness set for plural anaphora; the set-based condition (72) is GeneralWC_Incr with
IsEveryW.
Equations
- Cooper2023.SemUniversal restr scope = ((x : E) → restr x → scope x)
Instances For
no(P, Q) under its particular witness condition (Ch. 7, (70)): every witness of the
restrictor precludes the scope.
Equations
- Cooper2023.SemNo restr scope = Quantification.ParticularWC_No restr scope
Instances For
(92): a is a P, the copula over the indefinite article, is witnessed iff P(a) is, so
the compositional content and the construction-based content of (86)–(87) are distinct but
equivalent types.
(94c): a P is a, the quantifiers in the other order, is witnessed iff P(a) is as
well; only the construction expresses P(a) itself, and (89) A conductor is Dudamel is
odd.
A monotone increasing quantifier.
Equations
- Q.IsMonIncr = ∀ (P P' : Cooper2023.Ppty E) (a : (x : E) → P x → P' x), Nonempty (Q P) → Nonempty (Q P')
Instances For
A parametric property.
Equations
Instances For
The Dudamel fragment #
Dudamel is a conductor (82c), the existential quantifier under the copula, is witnessed by Dudamel's conducting, and Beethoven is a conductor is not.
The individuals.
Instances For
Equations
- Cooper2023.Dudamel.instDecidableEqInd x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- Cooper2023.Dudamel.instReprInd = { reprPrec := Cooper2023.Dudamel.instReprInd.repr }
is a conductor (81c).
Equations
Instances For
Dudamel is a conductor is true.
Equations
- Cooper2023.Dudamel.dudamelIsAConductor = { x := Cooper2023.Dudamel.Ind.dudamel, pWit := Cooper2023.Dudamel.Conductor.mk, qWit := { down := ⋯ } }
Instances For
Beethoven is a conductor is false.
Modality and intensionality without possible worlds (Ch. 6) #
A modal type system (§1.4.3.5, (54); §6.3) is a family of possibilities sharing their types but differing in which objects witness them; equivalence, subtyping, necessity and possibility are defined over all possibilities, (1), or over those in which the types occur, (2). Necessity and possibility in language are relativised, as in Kratzer's semantics, to a background type and a topos, a dependent type from situations to types standing in for the accessibility relation, (20)–(24). Intensionality replaces sets of worlds by types (§6.5): an attitude holds when the type of the agent's long-term memory, religious beliefs or desires matches its complement modulo relabelling, directly or through a point of view, (39)–(92).
Modal type systems (§6.3) #
A modal system of types (§1.4.3.5, (54)): a family of possibilities over shared types.
Equations
- Cooper2023.ModalSystem M Ty Obj = (M → Cooper2023.Possibility Ty Obj)
Instances For
The extension of T in the possibility p, (1a).
Instances For
T occurs in the type system of the possibility p.
Instances For
Restrictive equivalence (1a): the same extension in every possibility.
Instances For
Restrictive subtyping (1b).
Instances For
Restrictive necessity (1c): witnessed in every possibility.
Instances For
Restrictive possibility (1d): witnessed in some possibility.
Instances For
Inclusive equivalence (2a): the same extension wherever both types occur.
Equations
Instances For
Inclusive subtyping (2b).
Equations
Instances For
Inclusive necessity (2c): witnessed wherever the type occurs.
Instances For
Inclusive possibility (2d), as the book prints it.
Instances For
The restrictive notions entail the inclusive ones (§6.3).
Modality with topoi (§6.4) #
The witness conditions for nec and poss go through four versions; the last, (23)–(24),
takes a topos in place of Kratzer's ideal, and, as Cooper notes, has no counterpart of the
ordering source.
A topos (20): a dependent type from situations of a background type to types.
Equations
Instances For
Compatibility (17): something is of both types.
Equations
- Cooper2023.Compatible T₁ T₂ = Nonempty (T₁ × T₂)
Instances For
A witness of nec(T, B, τ) (23): a situation of the background type B, B a subtype
of the topos's domain, and the type the topos returns for it a subtype of T.
- sit : B
The situation.
- sub : B → τ.bg
The background type as a subtype of the topos's domain.
The type the topos returns as a subtype of
T.
Instances For
Mary should eat her broccoli (25)–(31) #
The base situation (26) has the broccoli on Mary's plate and Mary loving it; the deontic
topos (28a) sends a situation of a child with food on her plate to her eating it, the bouletic
topos (28b) a situation of a child loving some food to her eating it, and
nec([e:eat(m,b)], T_broc, τ) is witnessed by either, (29)–(30).
Equations
- Cooper2023.Dinner.instDecidableEqInd x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- Cooper2023.Dinner.instReprInd = { reprPrec := Cooper2023.Dinner.instReprInd.repr }
Equations
- One or more equations did not get rendered due to their size.
Instances For
The ptypes of the base situation (26), each witnessed by its fact.
- mk : Broccoli Ind.broccoli
Instances For
The deontic topos τ₁ (28a).
Equations
- Cooper2023.Dinner.deontic = { bg := Cooper2023.Dinner.OnPlate, fg := fun (r : Cooper2023.Dinner.OnPlate) => Cooper2023.Dinner.Eat r.y r.x }
Instances For
The bouletic topos τ₂ (28b).
Equations
- Cooper2023.Dinner.bouletic = { bg := Cooper2023.Dinner.Loves, fg := fun (r : Cooper2023.Dinner.Loves) => Cooper2023.Dinner.Eat r.y r.x }
Instances For
The base situation: the broccoli on Mary's plate, which she loves.
Equations
- One or more equations did not get rendered due to their size.
Instances For
(29a): eating the broccoli is necessary under the deontic topos, the base type a subtype of the topos's domain by (27) and the topos returning the type itself (30).
Equations
- One or more equations did not get rendered due to their size.
Instances For
(29b): and under the bouletic topos.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Intensionality (§6.5) #
Subtyping modulo relabelling (39), T₁ ⊑⇝ T₂.
Equations
- Cooper2023.RelabeledSubtype T₁ T₂ = Nonempty (T₁ → T₂)
Instances For
An agent's total information state (91), long-term memory, religious beliefs and desires as types, with the point-of-view relation on types, (55), (80).
- ltm : Agent → Type
The type of the agent's long-term memory.
- rbel : Agent → Type
The type of the agent's religious beliefs.
- des : Agent → Type
The type of the agent's desires.
pov M T:Mis a complete point of view onT, the asymmetric merge ofTwith an alternative type on some of its labels.
Instances For
An information type I matches T directly, or through a complete point of view M on
a type I matches, M matching T, (58), (80), (92).
Equations
- s.Matches I T = (Cooper2023.RelabeledSubtype I T ∨ ∃ (T₁ : Type) (M : Type), Cooper2023.RelabeledSubtype I T₁ ∧ s.pov M T₁ ∧ Cooper2023.RelabeledSubtype M T)
Instances For
Postulated subtyping: buying and selling, (35), (44)–(47), (50) #
A selling situation.
- seller : E
- thing : E
- buyer : E
Instances For
A buying situation.
- buyer : E
- thing : E
- seller : E
Instances For
The postulate (50b), sell(a, b, c) ⊑ buy(c, b, a), holding only in the possibilities the
postulate restricts attention to, unlike the structural (50a), BoyHugsDog.toBoyAndDog.
Instances For
believe(a, T) (40): the type of a's long-term memory matches T.
Equations
- Cooper2023.Believe s a T = Cooper2023.RelabeledSubtype (s.ltm a) T
Instances For
believe(a, T) with a point of view (58).
Equations
- Cooper2023.BelievePov s a T = s.Matches (s.ltm a) T
Instances For
rbelieve(a, T), characterised after (74): the type of a's religious beliefs matches
T.
Equations
- Cooper2023.RBelieve s a T = Cooper2023.RelabeledSubtype (s.rbel a) T
Instances For
want†(a, T) (92): a's desires match T.
Equations
- Cooper2023.WantDagger s a T = s.Matches (s.des a) T
Instances For
worship(a, Q) (75), (81): a's religious beliefs match the quantifier exported over
worship†, intentionality and specificity without existence.
Equations
- Cooper2023.Worship s dagger a Q = s.Matches (s.rbel a) (Q (dagger a))
Instances For
want_P(a, P) (90a): wanting to have a property.
Equations
- Cooper2023.WantP s a P = Cooper2023.WantDagger s a (P a)
Instances For
want_Q(a, Q) (90b): wanting a quantifier's worth of things is wanting to have them.
Equations
- Cooper2023.WantQ s have_ a Q = Cooper2023.WantDagger s a (Q (have_ a))
Instances For
Hesperus and Phosphorus, (52)–(53) #
The ancients' long-term memory (52): a body named Hesperus rising in the evening and a body named Phosphorus rising in the morning.
Instances For
After learning that they are one body (53): the manifest field, a subtype of (52) by the
projection toTwoStars.
Instances For
Intensional transitive verbs, (63)–(66), (87) #
A transitive verb whose predicate takes a quantifier (64), with the variant p† between
individuals.
The ptype of the verb over an individual and a quantifier.
- dagger : E → E → Type
The variant
p†over two individuals.
Instances For
(65): an extensional verb's ptype is equivalent to the quantifier exported over p†.
Equations
- v.IsExtensional = ∀ (a : E) (Q : Cooper2023.Quant E), Nonempty (v.pred a Q ≃ Q (v.dagger a))
Instances For
(66): a successful search is a finding.
The ptype of an event's success.
The subtyping
successful(seek(a, Q)) ⊑ find(a, Q).
Instances For
(87): booking a monotone increasing quantifier's worth of tables requires tables to be, without requiring a specific one.
Equations
- Cooper2023.BookRequiresBeing book be = ∀ (a : E) (Q : Cooper2023.Quant E), Q.IsMonIncr → Nonempty (book a Q) → Nonempty (Q be)
Instances For
Restrictive against inclusive necessity #
Two possibilities over the types rain and snow: snow is witnessed only in the first, so it
is possible but not necessary; and when snow does not occur in the second at all, it is
inclusively but not restrictively necessary, so the entailment of §6.3 does not reverse.
Equations
- Cooper2023.Weather.instDecidableEqTy x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- Cooper2023.Weather.instDecidableEqObj x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Both types occur in both possibilities; rain is witnessed in both, snow in the first.
Equations
- One or more equations did not get rendered due to their size.
Instances For
As system, but snow does not occur in the second possibility.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Witness-based quantification (Ch. 7) #
A property may be restricted by conditions in its domain beyond the required x-field,
(7b), and purification lowers the restriction into the body existentially, 𝔓 (12), or
universally, 𝔓∀ (13). The cardinality conditions on witness sets (20)–(35) have
frequentist probabilistic forms (41)–(58), estimable from an agent's experience base of
remembered judgements (37)–(40). The particular witness conditions for exist (63) and
no (70) are types equivalent to the general ones (59) whose witnesses carry what discourse
anaphora picks up.
A restricted property (7b): conditions on the individual in the domain, and the body.
- restr : E → Type
The restriction: the conditions the domain places on the individual.
The body: the type returned for an individual meeting the restriction.
Instances For
A property is pure (7a) when its restriction is trivial.
Instances For
Purification 𝔓(P) (12): the restriction lowered into the body under the local context.
Equations
- Cooper2023.Purify P x = ((c : P.restr x) × P.body x c)
Instances For
Universal purification 𝔓∀(P) (13): the body under every way of meeting the
restriction.
Equations
- Cooper2023.PurifyUniv P x = ((c : P.restr x) → P.body x c)
Instances For
Alignment of paths in the domain (Ch. 8, (51)–(52)): a manifest field identifying two paths is a further restriction of the domain, through which the body is read.
Instances For
Property restriction P|ℱ (Ch. 5, (98)): the domain narrowed by a property, the
alignment along the projection.
Equations
- P.restrictBy R = P.align (fun (x : E) => R x × P.restr x) fun (x : E) => Prod.snd
Instances For
For a pure property the two purifications agree: 𝔓 and 𝔓∀ differ only under a
non-trivial restriction.
Witness sets and probabilities (§7.3) #
The frequentist conditional probability (36) of one extension given another, 0 when
the condition is unwitnessed.
Equations
- Cooper2023.condProb A B = ↑(A ∩ B).card / ↑B.card
Instances For
(51)–(52): for a witness set X of objects with the property, the probability of 𝔗(X)
given 𝔗(P) is the proportion |X| / |[↓P]|.
(50): the probabilistic witness condition for most is the cardinal one (29).
An experience base (37): the judgements [sit = a, type = T] an agent remembers.
Equations
- Cooper2023.ExperienceBase E Ty = Finset (E × Ty)
Instances For
The extension of a type with respect to the experience base (38).
Equations
- 𝔍.extension T = Finset.image Prod.fst ({x ∈ 𝔍 | x.2 = T})
Instances For
The estimate p_𝔍(T₁ ‖ T₂) (39) of the probability (36) from the experience base.
Equations
- 𝔍.estimate T₁ T₂ = Cooper2023.condProb (𝔍.extension T₁) (𝔍.extension T₂)
Instances For
Witness conditions and anaphora (§7.4) #
With dog' and bark' the properties (61a–b), a witness for exist(dog', bark') under the
particular condition (63) is a dog that barks, whose x-field is what it picks up in
A dog is barking. It is right outside my window (64); under the particular condition for
no (70), No dog barked. They were all busy gnawing on a bone (71) has they pick up the
witness set of every dog, complement set anaphora; and under the general condition for
most (74), they in (75) picks up the witness set of most dogs. The properties are
decidable predicates lifted to types, as the set-based conditions require.
Equations
- Cooper2023.Dogs.instDecidableEqInd x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- Cooper2023.Dogs.instReprInd.repr Cooper2023.Dogs.Ind.fido prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "Cooper2023.Dogs.Ind.fido")).group prec✝
- Cooper2023.Dogs.instReprInd.repr Cooper2023.Dogs.Ind.rex prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "Cooper2023.Dogs.Ind.rex")).group prec✝
- Cooper2023.Dogs.instReprInd.repr Cooper2023.Dogs.Ind.spot prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "Cooper2023.Dogs.Ind.spot")).group prec✝
- Cooper2023.Dogs.instReprInd.repr Cooper2023.Dogs.Ind.luna prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "Cooper2023.Dogs.Ind.luna")).group prec✝
Instances For
Equations
- Cooper2023.Dogs.instReprInd = { reprPrec := Cooper2023.Dogs.instReprInd.repr }
Equations
- Cooper2023.Dogs.instFintypeInd = { elems := { val := ↑Cooper2023.Dogs.Ind.enumList, nodup := Cooper2023.Dogs.Ind.enumList_nodup }, complete := Cooper2023.Dogs.instFintypeInd._proof_1 }
Fido, Rex and Spot are dogs.
Equations
Instances For
Equations
- Cooper2023.Dogs.instDecidablePredIndIsDog x✝ = id inferInstance
Fido and Spot bark.
Equations
- Cooper2023.Dogs.Bark x = (x = Cooper2023.Dogs.Ind.fido ∨ x = Cooper2023.Dogs.Ind.spot)
Instances For
Equations
- Cooper2023.Dogs.instDecidablePredIndBark x✝ = id inferInstance
The property dog' (61a).
Equations
- Cooper2023.Dogs.dog x = PLift (Cooper2023.Dogs.IsDog x)
Instances For
The property bark' (61b).
Equations
- Cooper2023.Dogs.bark x = PLift (Cooper2023.Dogs.Bark x)
Instances For
A dog barks (63): Fido.
Equations
- Cooper2023.Dogs.aDogBarks = { x := Cooper2023.Dogs.Ind.fido, pWit := { down := Cooper2023.Dogs.aDogBarks._proof_1 }, qWit := { down := Cooper2023.Dogs.aDogBarks._proof_2 } }
Instances For
No dog barks is false: Fido is a dog that barks.
Most dogs bark (74): the witness set of Fido and Spot, more than half of the dogs, each barking, which they picks up in (75).
Equations
- One or more equations did not get rendered due to their size.
Instances For
(50): the same witness set by its probability, two thirds of the dogs.
Equations
- Cooper2023.Dogs.instDecidableEqTy x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- Cooper2023.Dogs.instReprTy = { reprPrec := Cooper2023.Dogs.instReprTy.repr }
Equations
- Cooper2023.Dogs.instReprTy.repr Cooper2023.Dogs.Ty.dog prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "Cooper2023.Dogs.Ty.dog")).group prec✝
- Cooper2023.Dogs.instReprTy.repr Cooper2023.Dogs.Ty.bark prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "Cooper2023.Dogs.Ty.bark")).group prec✝
Instances For
An experience base of three dogs, two of which were judged to bark.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The estimate p_𝔍(bark ‖ dog) is two thirds.
Type-based underspecification (Ch. 8) #
The content of an utterance is raised to a type of contents, the readings being the closure
of the compositional content under the operations of this section. Storage puts a
parametric quantifier into the context's store, leaving in its place the content of a label's
value (17), and retrieval quantifies it back in over the content as a property of that
value (19). Anaphoric combination identifies a pronoun's label with an antecedent's (28)
unless the pronoun is marked local, the marking cleared at the sentence boundary (77);
reflexives are marked (83), bound by reflexivisation (84) and required to be bound at the
verb phrase (85)–(88). Donkey anaphora goes through localisation (49), which folds the
context into the property's domain so that the indefinite's witness in the restrictor can be
aligned with the pronoun (51)–(52); 𝔓 then gives the weak reading (55)–(59) and 𝔓∀ the
strong one (60)–(66), quantifying over farmers and not farmer–donkey pairs.
Parametric contents and their context types (§8.2–8.3) #
A context type (§4.3, (16); §8.3, (82)): the labels a content requires the context to
assign to stored quantifiers, 𝔮, and to pronouns, 𝔰, and among the latter those marked
local, 𝔩, and reflexive, 𝔯.
- stored : Finset ℕ
The labels of stored quantifiers,
𝔮. - pronouns : Finset ℕ
The labels of pronouns,
𝔰. - locals : Finset ℕ
The labels of pronouns marked local,
𝔩. - reflexives : Finset ℕ
The labels of reflexives,
𝔯.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
The merge of two context types.
Equations
- One or more equations did not get rendered due to their size.
The context type requiring nothing.
Equations
- Cooper2023.CntxtType.instBot = { bot := { stored := ∅, pronouns := ∅, locals := ∅, reflexives := ∅ } }
A parametric content (14) over the contexts of Ch. 8, assignments of individuals to labels: its background context type and its foreground, a content for each assignment.
- bg : CntxtType
The background: the context type the content requires.
- fg : (ℕ → E) → C
The foreground: the content under each assignment.
Instances For
The content given a value for a label, the context specification c[𝔰.xᵢ = a] behind
retrieval (19), reflexivisation (84) and the alignment of a pronoun with its antecedent
(42)–(44).
Equations
Instances For
Storage (17): a stored quantifier leaves in its place the content of its label's value, required of the store and of the pronoun assignment.
Equations
- Cooper2023.Content.store i = { bg := { stored := {i}, pronouns := {i}, locals := ∅, reflexives := ∅ }, fg := fun (g : ℕ → E) => Cooper2023.SemPropName (g i) }
Instances For
A pronoun (75): the content of its label's value, marked local.
Equations
- Cooper2023.Content.pronoun i = { bg := { stored := ∅, pronouns := {i}, locals := {i}, reflexives := ∅ }, fg := fun (g : ℕ → E) => Cooper2023.SemPropName (g i) }
Instances For
A reflexive (83): the content of its label's value, marked reflexive.
Equations
- Cooper2023.Content.reflexive i = { bg := { stored := ∅, pronouns := {i}, locals := ∅, reflexives := {i} }, fg := fun (g : ℕ → E) => Cooper2023.SemPropName (g i) }
Instances For
The relabelling [α]𝔰.xⱼ ⇝ 𝔰.xᵢ: the content reads the label j as i, and the
background drops j.
Equations
Instances For
Anaphoric combination α @ᵢ,ⱼ β (28), (76) is defined when the functor requires the
label i and the argument requires j as a pronoun that is neither a stored quantifier's
nor marked local.
Equations
Instances For
Equations
- α.instDecidableAnaphoricDefined i j β = id inferInstance
The boundary operation B (77): the local marking is cleared at the sentence.
Equations
Instances For
Reflexivisation ℜ (84): the reflexive i is bound to the property's argument, its
label discharged and all reflexive marking cleared.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Principle A (85): a content with a reflexive still marked is excluded at the verb phrase (88).
Equations
- α.IsAnaphorFree = (α.bg.reflexives = ∅)
Instances For
Equations
- α.instDecidableIsAnaphorFree = id inferInstance
Every boy hugged a dog (§8.1, (1)) #
Two boys each hugging a different dog: the reading (1a) with the quantifiers in surface order is witnessed, and the reading (1b), the object quantifier stored and retrieved over the sentence, is not.
Equations
- Cooper2023.Hugging.instDecidableEqInd x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Tom and Bill are boys.
Equations
- Cooper2023.Hugging.Boy Cooper2023.Hugging.Ind.tom = PUnit.{1}
- Cooper2023.Hugging.Boy Cooper2023.Hugging.Ind.bill = PUnit.{1}
- Cooper2023.Hugging.Boy x✝ = Empty
Instances For
Fido and Rex are dogs.
Equations
- Cooper2023.Hugging.Dog Cooper2023.Hugging.Ind.fido = PUnit.{1}
- Cooper2023.Hugging.Dog Cooper2023.Hugging.Ind.rex = PUnit.{1}
- Cooper2023.Hugging.Dog x✝ = Empty
Instances For
Tom hugs Fido, Bill hugs Rex.
Equations
- Cooper2023.Hugging.Hug Cooper2023.Hugging.Ind.tom Cooper2023.Hugging.Ind.fido = PUnit.{1}
- Cooper2023.Hugging.Hug Cooper2023.Hugging.Ind.bill Cooper2023.Hugging.Ind.rex = PUnit.{1}
- Cooper2023.Hugging.Hug x✝¹ x✝ = Empty
Instances For
every boy, requiring nothing of the context.
Equations
- Cooper2023.Hugging.everyBoy = { bg := ⊥, fg := fun (x : ℕ → Cooper2023.Hugging.Ind) => Cooper2023.SemUniversal Cooper2023.Hugging.Boy }
Instances For
a dog, requiring nothing of the context.
Equations
- Cooper2023.Hugging.aDog = { bg := ⊥, fg := fun (x : ℕ → Cooper2023.Hugging.Ind) => Cooper2023.SemIndefArt Cooper2023.Hugging.Dog }
Instances For
(1a): the quantifiers in surface order.
Equations
Instances For
(1b): the object quantifier stored (17) and retrieved (19) over the sentence.
Equations
Instances For
Retrieval discharges what storage required.
(1a) is witnessed: every boy is such that there is a dog he hugged.
Equations
- Cooper2023.Hugging.surfaceWitness g Cooper2023.Hugging.Ind.tom x_2 = { x := Cooper2023.Hugging.Ind.fido, pWit := PUnit.unit, qWit := PUnit.unit }
- Cooper2023.Hugging.surfaceWitness g Cooper2023.Hugging.Ind.bill x_2 = { x := Cooper2023.Hugging.Ind.rex, pWit := PUnit.unit, qWit := PUnit.unit }
Instances For
(1b) is not: there is no dog every boy hugged.
Localisation and donkey anaphora (§8.3) #
Localisation ℒ (49): the context a parametric property requires is folded into the
property's domain under the label 𝔠, giving a restricted property.
Equations
Instances For
No dog which chases a cat catches it (46a) #
The scope is the localised catches it restricted by the restrictor and aligned so that the
caught cat is the chased one (50)–(51); under the particular condition for no the sentence
(55) says that every dog which chases a cat fails to be a dog which chases a cat and catches
it.
Equations
- Cooper2023.Chasing.instDecidableEqInd x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
The dogs.
Equations
- Cooper2023.Chasing.Dog Cooper2023.Chasing.Ind.dog₁ = PUnit.{1}
- Cooper2023.Chasing.Dog Cooper2023.Chasing.Ind.dog₂ = PUnit.{1}
- Cooper2023.Chasing.Dog x✝ = Empty
Instances For
The cats.
Equations
- Cooper2023.Chasing.Cat Cooper2023.Chasing.Ind.cat₁ = PUnit.{1}
- Cooper2023.Chasing.Cat Cooper2023.Chasing.Ind.cat₂ = PUnit.{1}
- Cooper2023.Chasing.Cat x✝ = Empty
Instances For
Each dog chases one cat.
Equations
Instances For
catches it (47): the pronoun's referent supplied by the context.
Equations
- Cooper2023.Chasing.catchesIt Catch = { bg := Cooper2023.Chasing.Ind, fg := fun (y x : Cooper2023.Chasing.Ind) => Catch x y }
Instances For
dog which chases a cat, the restrictor, as the domain of (50): a dog with a cat it chases.
Equations
Instances For
The scope (51): catches it localised, restricted by the restrictor and aligned so that
it is the chased cat.
Equations
- One or more equations did not get rendered due to their size.
Instances For
(55): no(restr, scope) with the scope purified.
Equations
Instances For
True when no dog catches anything.
Equations
- One or more equations did not get rendered due to their size.
Instances For
False when every dog catches the cat it chases.
Every farmer who owns a donkey likes it (58)–(66) #
The localised likes it restricted by farmer who owns a donkey and aligned (65) is the
property of being a farmer who owns a donkey and likes that donkey; its purification 𝔓
gives the weak reading, some donkey she owns, and 𝔓∀ (66) the strong one, every donkey
she owns, the readings of [Kan94]. A farmer who owns two donkeys and likes one
separates them.
Equations
- Cooper2023.Donkeys.instDecidableEqInd x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
The farmers.
Equations
- Cooper2023.Donkeys.Farmer Cooper2023.Donkeys.Ind.farmer₁ = PUnit.{1}
- Cooper2023.Donkeys.Farmer Cooper2023.Donkeys.Ind.farmer₂ = PUnit.{1}
- Cooper2023.Donkeys.Farmer x✝ = Empty
Instances For
The donkeys.
Equations
- Cooper2023.Donkeys.Donkey Cooper2023.Donkeys.Ind.donkey₁ = PUnit.{1}
- Cooper2023.Donkeys.Donkey Cooper2023.Donkeys.Ind.donkey₂ = PUnit.{1}
- Cooper2023.Donkeys.Donkey x✝ = Empty
Instances For
The first farmer owns both donkeys, the second the second.
Equations
- Cooper2023.Donkeys.Own Cooper2023.Donkeys.Ind.farmer₁ Cooper2023.Donkeys.Ind.donkey₁ = PUnit.{1}
- Cooper2023.Donkeys.Own Cooper2023.Donkeys.Ind.farmer₁ Cooper2023.Donkeys.Ind.donkey₂ = PUnit.{1}
- Cooper2023.Donkeys.Own Cooper2023.Donkeys.Ind.farmer₂ Cooper2023.Donkeys.Ind.donkey₂ = PUnit.{1}
- Cooper2023.Donkeys.Own x✝¹ x✝ = Empty
Instances For
Each farmer likes one donkey.
Equations
Instances For
farmer who owns a donkey.
Equations
Instances For
likes it (61), localised (62)–(63), restricted (64) and aligned (65).
Equations
- One or more equations did not get rendered due to their size.
Instances For
The weak reading (59): every farmer who owns a donkey likes some donkey she owns.
Equations
- Cooper2023.Donkeys.weak Cooper2023.Donkeys.Ind.farmer₁ x_2 = ⟨(PUnit.unit, ⟨Cooper2023.Donkeys.Ind.donkey₁, (PUnit.unit, PUnit.unit)⟩), PUnit.unit⟩
- Cooper2023.Donkeys.weak Cooper2023.Donkeys.Ind.farmer₂ x_2 = ⟨(PUnit.unit, ⟨Cooper2023.Donkeys.Ind.donkey₂, (PUnit.unit, PUnit.unit)⟩), PUnit.unit⟩
Instances For
The strong reading (60), (66) fails: the first farmer does not like the second donkey.
Pronouns, locality and reflexives (30)–(36), (67)–(88) #
With the subject stored, its label is available for anaphoric combination with the pronoun
of thinks she failed, whose local marking the embedded sentence's boundary has cleared;
retrieval then quantifies over the property of being a girl who thinks she failed, (36c).
In Sam likes him the pronoun is still marked local when the subject combines, so the
combination is undefined, Principle B; likes himself reflexivised is like(x, x), its
marking cleared for the verb phrase's filter, which excludes the reflexive left unbound,
Principle A.
Equations
- Cooper2023.Binding.instDecidableEqInd x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Ann is the girl.
Equations
- Cooper2023.Binding.Girl Cooper2023.Binding.Ind.ann = PUnit.{1}
- Cooper2023.Binding.Girl x✝ = Empty
Instances For
think(x, T): a thought with the type's witness.
- thought : T
Instances For
Sam (70), requiring nothing of the context.
Equations
- Cooper2023.Binding.sam = { bg := ⊥, fg := fun (x : ℕ → Cooper2023.Binding.Ind) => Cooper2023.SemPropName Cooper2023.Binding.Ind.sam }
Instances For
no girl (33).
Equations
- Cooper2023.Binding.noGirl = { bg := ⊥, fg := fun (x : ℕ → Cooper2023.Binding.Ind) => Cooper2023.SemNo Cooper2023.Binding.Girl }
Instances For
failed.
Equations
- Cooper2023.Binding.failed = { bg := ⊥, fg := fun (x : ℕ → Cooper2023.Binding.Ind) => Cooper2023.Binding.Fail }
Instances For
thinks, over a ptype of thinking.
Equations
- Cooper2023.Binding.thinks = { bg := ⊥, fg := fun (x : ℕ → Cooper2023.Binding.Ind) (T : Type) (x_1 : Cooper2023.Binding.Ind) => Cooper2023.Binding.Think x_1 T }
Instances For
thinks she failed (31): she failed a sentence, its pronoun no longer local past the boundary.
Equations
Instances For
The stored subject (34) combines anaphorically with the pronoun, (35).
Without the boundary of the embedded sentence the pronoun would still be local.
No girl thinks she failed (36): the stored no girl retrieved over the anaphoric
combination is no(girl', λx. think(x, fail(x))).
likes him (69), the pronoun marked local.
Instances For
(72)–(73): him cannot be related to the stored Sam (71) within the clause, Principle B.
likes himself reflexivised (84).
Equations
Instances For
(68c): the reflexivised property is like(x, x).
Reflexivisation clears the marking for the verb phrase's filter (88).
The filter excludes the reflexive left unbound.
Sam likes himself (73), the stored Sam retrieved, is witnessed by Sam's liking himself.
Instances For
A man walked. He whistled. (37)–(44) #
The pronoun's label is identified with the man of the previous utterance's witness (42) and the dependency on the label replaced by one on the man (43)–(44).
Equations
- Cooper2023.Whistling.instDecidableEqInd x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
John is a man.
Equations
Instances For
John walks.
Equations
Instances For
John whistles.
Equations
Instances For
The content of a man walked under the particular condition for exist (38).
Equations
Instances For
John walked.
Equations
- Cooper2023.Whistling.aManWalked = { x := Cooper2023.Whistling.Ind.john, pWit := PUnit.unit, qWit := PUnit.unit }
Instances For
he whistled (39): the pronoun's label required of the context.
Equations
- Cooper2023.Whistling.heWhistled = ((Cooper2023.Content.pronoun 0).app { bg := ⊥, fg := fun (x : ℕ → Cooper2023.Whistling.Ind) => Cooper2023.Whistling.Whistle }).boundary
Instances For
(42)–(44): given the previous utterance's witness, the pronoun's label is no longer required and the content is that the man whistled.
The man of the previous utterance whistled.
Equations
- Cooper2023.Whistling.heWhistledWitness g = PUnit.unit
Instances For
The book's examples #
A selection of the English examples of Chs. 3, 6, 7 and 8 are the rows of
Data/Examples/Cooper2023.json. The rows with a quantifier feature are the
discourse-anaphora examples of §7.4, §7.4.1 and §8.3, each reading named by the anaphora
set the pronoun picks up.
The quantifier relation a row's determiner names.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The anaphora set a reading names.
Equations
- Cooper2023.anaphoraRefs = [("refset", Quantification.AnaphoraRef.refset), ("maxset", Quantification.AnaphoraRef.maxset), ("compset", Quantification.AnaphoraRef.compset)]
Instances For
A reading of a pronoun with a quantified antecedent is acceptable exactly when a witness for the content provides a path to its anaphora set.