You have 3 free guides left 😟
Unlock your guides
You have 3 free guides left 😟
Unlock your guides

11.3 Discourse representation and centering theory

2 min readjuly 19, 2024

models meaning across sentences, building structures with referents and . It interprets sentences based on context and resolves by linking to referents. This approach captures how meaning accumulates through discourse.

tracks entity in discourse, classifying transitions between utterances. It aids anaphora resolution by preferring the most salient entity as antecedent. Both theories offer complementary insights into how we interpret connected language.

Discourse Representation Theory

Concepts of discourse representation theory

Top images from around the web for Concepts of discourse representation theory
Top images from around the web for Concepts of discourse representation theory
  • Models meaning of a discourse (sequence of sentences or utterances in a context)
  • Constructs a (DRS) captures cumulative meaning of a discourse
    • DRS consists of (entities introduced in the discourse, represented by variables like x, y) and conditions (predicate-argument structures or relations between discourse referents)
  • Assumes interpretation of a sentence depends on context provided by preceding discourse
  • Resolves anaphoric expressions (pronouns) by linking them to appropriate discourse referents in the DRS

Centering Theory and Anaphora Resolution

Application of centering theory

  • Discourse model tracks local attentional state (salience of entities) in a discourse
  • Each utterance evokes forward-looking centers (Cf, potential candidates for Cb of following utterance, ranked by salience) and a single (Cb, most salient entity in previous utterance realized in current utterance)
  • Classifies transitions between utterances based on Cb and Cf relationships
    1. Continue: Cb(Un) = Cb(Un-1), Cb(Un) is most salient entity in Cf(Un)
    2. Retain: Cb(Un) = Cb(Un-1), Cb(Un) is not most salient entity in Cf(Un)
    3. Smooth-shift: Cb(Un) ≠ Cb(Un-1), Cb(Un) is most salient entity in Cf(Un)
    4. Rough-shift: Cb(Un) ≠ Cb(Un-1), Cb(Un) is not most salient entity in Cf(Un)
  • Predicts discourses with more continue transitions are more coherent

Discourse structure in anaphora resolution

  • (pronouns, definite descriptions) depend on discourse structure for interpretation
  • DRT resolves anaphoric references by linking them to accessible discourse referents in DRS
    • determined by of DRS and scope of and
  • Centering theory resolves anaphoric references by preferring Cb as antecedent
    • If Cb is not suitable antecedent, considers other entities in Cf in order of salience

Discourse representation vs centering theory

  • Both DRT and centering theory aim to resolve anaphoric references in discourse
  • DRT focuses on global discourse structure and accessibility of discourse referents
    • Constructs of discourse (DRS) to resolve anaphora
  • Centering theory focuses on local attentional state and salience of entities
    • Uses Cb and Cf to resolve anaphora based on
  • DRT provides more comprehensive account of anaphora resolution (considers quantifier scope, negation)
  • Centering theory offers insights into role of entity salience and local coherence in anaphora resolution
  • Approaches can be seen as complementary (DRT captures global discourse structure, centering theory captures local attentional state)
© 2024 Fiveable Inc. All rights reserved.
AP® and SAT® are trademarks registered by the College Board, which is not affiliated with, and does not endorse this website.


© 2024 Fiveable Inc. All rights reserved.
AP® and SAT® are trademarks registered by the College Board, which is not affiliated with, and does not endorse this website.

© 2024 Fiveable Inc. All rights reserved.
AP® and SAT® are trademarks registered by the College Board, which is not affiliated with, and does not endorse this website.
Glossary
Glossary