← Back to catalogue
Published

coreference

vr.tr.coreference · thing-coreference

Let an agent identify and resolve coreference in text, explain the phenomenon and its types, avoid ambiguity in writing, and support language processing and teaching.

Thing Registry Cross-cutting context XCT.REL

Bundle → Layer → Finding → Questions Filled

4 bundles · 8 layers · 8 findings · 16 questions

Resolve Finding referents.

Pronoun

Pronoun resolution.

Pronoun

Resolution.

  1. To what does this pronoun or phrase refer in the text? definition
  2. Is the reference ambiguous? boundary

Chains

Entity chains.

Chains

Chains.

  1. Which expressions in the text refer to the same entity? definition
  2. How many distinct entities are mentioned? measurement
Write Clear writing.

Clarity

Ambiguous references.

Clarity

Clarity.

  1. Could a reader misread which entity this pronoun refers to? boundary
  2. How can the sentence be rewritten clearly? action

Style

Reference style.

Style

Style.

  1. When should a name be repeated rather than a pronoun used? action
  2. How do style guides advise? provenance
Theory Linguistics.

Types

Kinds of relation.

Types

Types.

  1. Is this anaphora, cataphora, bridging or exophora? definition
  2. How do binding constraints restrict coreference? definition

Languages

Cross-linguistic.

Languages

Languages.

  1. How does coreference work in this language, such as pro-drop or gendered pronouns? provenance
  2. Which challenges arise in translation? provenance
NLP Language processing.

Systems

Coreference resolution systems.

Systems

Systems.

  1. How do coreference resolution systems work, and how are they evaluated? provenance
  2. Which datasets and metrics are standard? provenance

Annotate

Annotation.

Annotate

Annotation.

  1. How should coreference be annotated in this corpus? action
  2. Which scheme applies? provenance

Classifiers Filled

Family
Thing Registry
Category
Cross-cutting context
Entry kind
noun.relation
Plane
XCT
Domain
XCT.REL
Facets: substance
abstract
Facets: origin
conceptual
Facets: agency
inert
Facets: mobility
not-applicable
Facets: scale
not-applicable
Facets: affordances
observable

What it is Filled

The grammatical relation in which two or more expressions in a text refer to the same entity, as when a pronoun refers back to a noun phrase or two names denote the same person; resolving coreference is essential to understanding text and is a core task in linguistics and natural language processing.

Why it exists Filled

Let an agent identify and resolve coreference in text, explain the phenomenon and its types, avoid ambiguity in writing, and support language processing and teaching.

Distinguishing features Filled

  • Same referent
  • Across expressions
  • Anaphoric or cataphoric
  • Key to comprehension

What robots and AI may and may not do Filled

Must not

  • Link mentions to a real person without evidence and present the link as fact.
  • Use coreference resolution to identify people in anonymized texts.
  • Assume a person's gender from a name when resolving pronouns, and present it as certain.

Only with a human decision

  • Using resolved identities in decisions about people, such as in legal or investigative records.

May

  • Resolve coreference in texts the user provides, stating uncertain links.
  • Explain how pronouns and names refer to entities in a text.

Moral aspects Filled

  • Wrongly linking mentions can attribute statements or acts to the wrong person.
  • Automatic resolution often misgenders people.

Who is affected

  • People mentioned in texts
  • Readers relying on the analysis

Owners Filled

Steward

Nobody: a linguistic relation held in common.

Links to other meta-models Filled

parent

  • oewn-13818991-n - registry parent class

related

  • grammatical relation - category
  • modification - other grammatical relation
  • natural language processing - field
  • unrelatedness - contrasting relation

What else AI and robots need to interact with it Filled

Identity and identifiers required Filled

  • Vercy registry: vr.tr.coreference
  • Open English WordNet: oewn-13821275-n

Direct properties not applicable Not applicable

Not applicable

Plane XCT: no invented physical properties.

Recognition optional Filled

  • Pronouns referring to earlier phrases
  • Repeated mentions of the same entity
  • Modification is a different grammatical relation
  • Not physical; links between expressions in text.

Capabilities and actions required Filled

  • resolve pronouns and referring expressions
  • identify ambiguous references
  • explain anaphora, cataphora and related relations
  • support coreference annotation

Hazards and failure modes required Filled

  • Ambiguous pronouns in writing
  • Wrong resolution in translation and summarisation
  • Confusing coreference with other relations

Standards and interfaces required Filled

  • Annotation schemes such as OntoNotes and the Universal Anaphora initiative

Context of use required Filled

  • Understanding who or what a text refers to.
  • pronominal coreference
  • nominal coreference between names and descriptions
  • anaphora and cataphora
  • bridging and near-identity relations
  • coreference chains in discourse

Sources Filled

  1. Open English WordNet 2024, synset oewn-13821275-n - sense and gloss of the concept
  2. Wiktionary: coreference - word senses and usage

Open questions

  • Should anaphora be a separate entry?
  • How should annotation resources be linked?
  • How should language differences be recorded?

Machine files

Provenance

thing registry research (pass 2) · unreviewed

Built from: models/things/publications/thing-coreference/spec.json