Documentation

Linglib.Data.Generalizations.HomogeneityGap

Generalizations.HomogeneityGap — cross-paper prediction target #

The homogeneity gap of unembedded plural definites (and their modal analogues): a positive sentence is true in the ALL scenario, its negation true in the NONE scenario, and in mixed (GAP) scenarios both are neither clearly true nor clearly false. The generalisation predates any single account ([Lob00], [Kri15b], [KC15], [AJ22]); rival explanations include double strengthening ([Mag14]), exhaustification ([BL21]), and trivalent supervaluation ([Kri16]).

Sibling of Generalizations.HomogeneityProjection, which covers the embedded cells (operator × scenario); this file covers the unembedded polarity × scenario grid. The Predict signatures differ (Polarity vs EmbeddingOperator first argument), so the two pools stay separate.

Main declarations #

Divergence theorems between rival accounts live in the comparing paper's study file, not here.

Substrate #

source
inductive Generalizations.HomogeneityGap.GapScenario :
Type

Scenario triad for unembedded homogeneity items: all-positive baseline, all-negative baseline, and the mixed scenario where the gap appears.

Instances For
    source
    @[implicit_reducible]
    instance Generalizations.HomogeneityGap.instReprGapScenario :
    Repr GapScenario
    Equations
    source
    def Generalizations.HomogeneityGap.instReprGapScenario.repr :
    GapScenario → ℕ → Std.Format
    Equations
    • One or more equations did not get rendered due to their size.
    Instances For
      source
      @[implicit_reducible]
      instance Generalizations.HomogeneityGap.instDecidableEqGapScenario :
      DecidableEq GapScenario
      Equations

      Test-suite schema #

      source
      @[reducible, inline]
      abbrev Generalizations.HomogeneityGap.GapPredict :
      Type

      Prediction signature for accounts of the unembedded homogeneity gap: given the sentence polarity and the scenario, the trivalent value the account assigns.

      Equations
      Instances For
        source
        structure Generalizations.HomogeneityGap.GapDatum :
        Type

        Empirical datum lifted from a paper-anchored LinguisticExample: observed is the trivalent value the paper's judgments commit to in this (polarity, scenario) cell.

        Instances For
          source
          @[implicit_reducible]
          instance Generalizations.HomogeneityGap.instReprGapDatum :
          Repr GapDatum
          Equations
          source
          def Generalizations.HomogeneityGap.instReprGapDatum.repr :
          GapDatum → ℕ → Std.Format
          Equations
          • One or more equations did not get rendered due to their size.
          Instances For
            source
            def Generalizations.HomogeneityGap.instDecidableEqGapDatum.decEq (x✝ x✝¹ : GapDatum) :
            Decidable (x✝ = x✝¹)
            Equations
            • One or more equations did not get rendered due to their size.
            Instances For
              source
              @[implicit_reducible]
              instance Generalizations.HomogeneityGap.instDecidableEqGapDatum :
              DecidableEq GapDatum
              Equations

              LinguisticExample adapter #

              source
              def Generalizations.HomogeneityGap.parsePolarity :
              String → Option Features.Polarity

              Read a Polarity from the paperFeatures "polarity" value.

              Equations
              Instances For
                source
                def Generalizations.HomogeneityGap.parseScenario :
                String → Option GapScenario

                Read a GapScenario from the paperFeatures "condition" value.

                Equations
                Instances For
                  source
                  def Generalizations.HomogeneityGap.baselineTruth (p : Features.Polarity) (s : GapScenario) :
                  Core.Duality.Truth3

                  Observed trivalent value for a baseline cell, determined by polarity: a positive sentence is true in ALL and false in NONE; a negative one the reverse.

                  Equations
                  Instances For
                    source
                    def Generalizations.HomogeneityGap.gapTruth (features : List (String × String)) :
                    Option Core.Duality.Truth3

                    Observed value for a GAP-scenario row. .indet when the paper detected the gap (gap_detected = "true"); otherwise the row must assert its bivalent judgment explicitly via a classical_value key (e.g. the gap-free strong-necessity cells of [AJ22]). Gap rows with neither key are not pool cells — they carry study-local refinements (removers, borderline-response items, issue-relativized judgments) and are excluded rather than assigned a fabricated value.

                    Equations
                    • One or more equations did not get rendered due to their size.
                    Instances For
                      source
                      def Generalizations.HomogeneityGap.fromExample (e : Data.Examples.LinguisticExample) :
                      Option GapDatum

                      Lift a LinguisticExample to a GapDatum via the polarity, condition, gap_detected, and classical_value keys. Rows tagged with an embedding key other than "unembedded" return none — embedded cells belong to Generalizations.HomogeneityProjection.

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

                        Pool #

                        source
                        def Generalizations.HomogeneityGap.allData :
                        List GapDatum

                        Cross-paper pool of unembedded homogeneity-gap data. Rival accounts pass their GapPredict implementations against this pool; per-datum and divergence theorems live in the comparing papers' study files.

                        Equations
                        Instances For