Generating Pragmatically Appropriate Sentences from Logic: The Case of the Conditional and Biconditional

Renhao Pei*, Kees van Deemter

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

Abstract

It is widely assumed that there exist mismatches between the connectives of Propositional Logic and their counterparts in Natural Language. One mismatch that has been extensively discussed is Conditional Perfection, the phenomenon in which a conditional sentence is interpreted as a biconditional under some circumstances. The Pragmatics literature has provided valuable insights into the question of whether Conditional Perfection will happen in a given context. In order to make these insights more explicit and testable, we designed an algorithm to generate pragmatically more appropriate sentences from propositional logical formulas involving material implication and biconditional implication. This algorithm was tested in an evaluation by human participants, in which generated sentences are compared against those generated by a simple baseline algorithm. The evaluation results suggest that the designed algorithm generates better sentences, which capture the semantics of the logical formulas more faithfully.
Original languageEnglish
Title of host publicationLogic and Algorithms in Computational Linguistics, LACompLing 2021
EditorsRoussanka Loukanova, Peter LeFanu Lumsdaine, Reinhard Muskens
Place of PublicationCham
PublisherSpringer
Pages171-190
Number of pages20
ISBN (Print)9783031217791
DOIs
Publication statusPublished - 12 Mar 2023

Publication series

NameStudies in Computational Intelligence
Volume1081
ISSN (Print)1860-949X
ISSN (Electronic)1860-9503

Keywords

  • Conditional perfection
  • Propositional logic
  • Natural language generation

Fingerprint

Dive into the research topics of 'Generating Pragmatically Appropriate Sentences from Logic: The Case of the Conditional and Biconditional'. Together they form a unique fingerprint.

Cite this