Centering — Information-Status Cf Ranking #
@cite{strube-hahn-1999} @cite{poesio-stevenson-eugenio-hitzeman-2004} @cite{prince-1981} @cite{gundel-hedberg-zacharski-1993}
@cite{strube-hahn-1999} (their §4.2) propose ranking forward-looking centers by information status (IS) — a non-grammatical-function ranker originally argued for free-word-order languages, in particular German. Their ranking projects @cite{prince-1981}'s familiarity scale onto three tiers:
HEARER-OLD ≺ MEDIATED ≺ HEARER-NEW
with linear-order tiebreak within each tier. The relation ≺ is
"more salient", so HEARER-OLD entities receive the highest Centering
rank; this file inverts the symbol and ranks HEARER-OLD = 2 / MEDIATED
= 1 / HEARER-NEW = 0.
@cite{poesio-stevenson-eugenio-hitzeman-2004} (henceforth PSDH) re-evaluate Strube-Hahn's ranking on English corpus data (the GNOME museum + pharmaceutical subcorpora; PSDH §3.2). Internally HEARER-OLD subsumes Prince's evoked + unused entities (e.g. Margaret Thatcher in a politically-knowledgeable context — presumed shared knowledge); MEDIATED covers Prince's inferable, containing- inferable, and anchored-brand-new (PSDH §2.4.3 fn 10); HEARER-NEW covers brand-new entities.
Empirical headline (PSDH §5.1.1). With STRUBE-HAHN ranking, both the IS and IF instantiations verify all three claims (Strong C1, Rule 1 GJW95, Rule 2 BFP) at the .01 level. Switching to GFTHERELIN (grammatical function + linear-order tiebreak), the same IS instantiation verifies Rule 2 only at the .05 level (PSDH §4.4.1 end). The .01-vs-.05 elevation is what STRUBE-HAHN ranking buys — not exclusivity over IF.
Substrate source for IS. The ofGivenness projection below
maps @cite{gundel-hedberg-zacharski-1993}'s 6-tier GivennessStatus
(in Features/Givenness.lean) onto Strube-Hahn's 3-tier IS — the
substantive linguistic bridge the two literatures imply. GHZ's
uniquelyIdentifiable/referential populate the MEDIATED tier
(Prince's inferable / containing-inferable / anchored-brand-new),
witnessed by mediated_reachable below.
@cite{strube-hahn-1999}'s information-status taxonomy as a
Centering-instance-private enum. The 3-tier shape (HEARER-OLD /
MEDIATED / HEARER-NEW) is preserved as the Strube-Hahn-native
surface; the substrate source is GHZ-6 (see ofGivenness below).
Strube-Hahn's HEARER-OLD covers @cite{prince-1981}'s discourse-old
- unused/evoked entities (e.g. Margaret Thatcher in a UK political context — presumed shared knowledge). MEDIATED covers Prince's inferable, containing-inferable, and anchored-brand-new. HEARER-NEW covers brand-new entities. Per @cite{poesio-stevenson-eugenio-hitzeman-2004} §2.4.3 (their footnote 10 spells out the Prince-correspondence).
- hearerOld : StrubeHahnInfoStatus
Hearer-old: discourse-old or unused/evoked. Highest salience.
- mediated : StrubeHahnInfoStatus
Mediated: inferable, containing-inferable, or anchored-brand-new (@cite{prince-1981}'s middle category).
- hearerNew : StrubeHahnInfoStatus
Hearer-new: brand-new entity. Lowest salience.
Instances For
Equations
- Discourse.Centering.instDecidableEqStrubeHahnInfoStatus x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
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.
Strube-Hahn IS rank on Nat: HEARER-OLD = 2 (highest Centering
rank, matching ≺ = more salient), MEDIATED = 1, HEARER-NEW = 0.
Equations
Instances For
The Strube-Hahn Cf-ranker instance — sibling of GrammaticalRole,
enabling the parametric Centering story PSDH §4.4 evaluate.
Total order via the rank pullback, mirroring the LinearOrder GrammaticalRole instance in the sibling file.
Equations
- One or more equations did not get rendered due to their size.
Project @cite{gundel-hedberg-zacharski-1993}'s 6-tier
Features.GivennessStatus onto Strube-Hahn's 3-tier IS, following
the GHZ→Prince→Strube-Hahn chain documented in
@cite{poesio-stevenson-eugenio-hitzeman-2004} §2.4.3 fn 10:
inFocus,activated,familiar→hearerOld(Prince's Evoked + Unused: currently activated, in working memory, or in long-term memory). Strube-Hahn's HEARER-OLD includes Prince's Unused tier — entities like Margaret Thatcher that are part of shared knowledge but not currently activated.uniquelyIdentifiable,referential→mediated(Prince's Inferable + Containing-Inferable + Anchored Brand-new — entities requiring inference or anchoring to be identified).typeIdentifiable→hearerNew(Prince's bare Brand-new).
The 3-2-1 split is the linguistically-defensible cut implied by
Strube-Hahn's own definition. The previous 2-2-2 implementation
here mistakenly pushed GHZ's familiar into MEDIATED. The
MEDIATED tier has a substrate source (the prior projection from
DiscourseStatus had no source for it, leaving MEDIATED
unreachable).
Equations
- Discourse.Centering.StrubeHahnInfoStatus.ofGivenness Features.GivennessStatus.inFocus = Discourse.Centering.StrubeHahnInfoStatus.hearerOld
- Discourse.Centering.StrubeHahnInfoStatus.ofGivenness Features.GivennessStatus.activated = Discourse.Centering.StrubeHahnInfoStatus.hearerOld
- Discourse.Centering.StrubeHahnInfoStatus.ofGivenness Features.GivennessStatus.familiar = Discourse.Centering.StrubeHahnInfoStatus.hearerOld
- Discourse.Centering.StrubeHahnInfoStatus.ofGivenness Features.GivennessStatus.uniquelyIdentifiable = Discourse.Centering.StrubeHahnInfoStatus.mediated
- Discourse.Centering.StrubeHahnInfoStatus.ofGivenness Features.GivennessStatus.referential = Discourse.Centering.StrubeHahnInfoStatus.mediated
- Discourse.Centering.StrubeHahnInfoStatus.ofGivenness Features.GivennessStatus.typeIdentifiable = Discourse.Centering.StrubeHahnInfoStatus.hearerNew
Instances For
MEDIATED-tier dissolution: substrate-anchor examples #
The headline of the post-Krifka substrate refactor: GHZ-6's
uniquelyIdentifiable and referential finally supply the MEDIATED
tier that the old DiscourseStatus-based projection couldn't reach.
These rfl examples lock the dissolution as grep-anchored facts.
MEDIATED is reachable: there exists a GivennessStatus that
projects to it (in fact two — uniquelyIdentifiable and
referential). Pre-Krifka the projection was from DiscourseStatus,
which had no MEDIATED-suitable input.
The projection respects the salience ordering: more cognitively activated givenness statuses map to higher (or equal) Centering ranks. Sanity check on the substantive linguistic claim that the GHZ-6 hierarchy and Strube-Hahn's IS scale agree about which referents are more salient.