Discourse Representation Theory (DRT) is a dynamic approach to understanding how meaning is built in conversations. It uses mental structures called DRSs to show how we process and connect ideas as we talk, handling things like pronouns and time references.

DRT goes beyond looking at single sentences, focusing on how meaning grows as we speak. It's especially good at dealing with context-dependent stuff like pronouns and assumptions, making it a powerful tool for studying how we make sense of language in real conversations.

Discourse Representation Theory (DRT)

Components of Discourse Representation Theory

• DRT dynamically models incremental meaning construction in discourse
• Key components:
  • Discourse Representation Structures (DRSs) mentally represent the semantic content of a discourse containing referents (discourse entities) and conditions (predicates and relations)
  • Construction algorithm incrementally builds and updates DRSs as discourse unfolds handling anaphora, presuppositions, and temporal relations
• Principles:
  • Noun phrases introduce discourse referents accessible for anaphoric reference
  • Conditions express properties and relations of discourse referents
  • DRSs dynamically update as new sentences are processed enabling context-dependent interpretation

Construction of discourse representation structures

• Basic construction:
  • Introduce new discourse referents for noun phrases
  • Add conditions representing predicates and relations in the sentence
• Example: "A man walks. He whistles."
  • DRS for first sentence:
    • [x]
    • [man(x), walk(x)]
  • Updated DRS after second sentence:
    • [x]
    • [man(x), walk(x), whistle(x)]
• Quantifiers:
  • Universal quantifiers introduce a new DRS box with restricted referents
  • Existential quantifiers introduce new referents in current DRS box
• Negation and disjunction:
  • Negation introduces a new DRS box with negated condition
  • Disjunction introduces two alternative DRS boxes

Semantic content in DRSs

• Anaphora resolution:
  • DRSs represent anaphoric dependencies
  • Pronouns refer to accessible discourse referents introduced earlier in DRS
• Presuppositions:
  • DRT treats presuppositions as conditions that must be satisfied in context for felicitous discourse
  • Presuppositions are separate DRS boxes merged with main DRS if satisfied
• Temporal relations:
  • DRT represents temporal ordering of events using additional conditions and discourse referents for time intervals
  • Temporal connectives ("before", "after") introduce relations between time intervals of described events

DRT vs other semantic theories

• First-order logic (FOL):
  • FOL statically assigns truth conditions to isolated sentences
  • DRT extends FOL with dynamic perspective handling discourse-level phenomena
• Montague semantics:
  • Montague semantics compositionally assigns meanings to sentences based on parts
  • DRT incorporates Montague semantics insights but focuses on incremental meaning construction in discourse
• Advantages of DRT:
  • More natural and intuitive representation of discourse structure and context-dependent meaning
  • Better handles anaphora, presuppositions, and temporal relations than static theories (FOL)
• Limitations of DRT:
  • Complex formalism requiring construction algorithm
  • Challenges handling complex linguistic phenomena (bridging anaphora, discourse relations)