Generating Definite Descriptions: Non incrementality, inference and data
Abstract
This paper presents two extensions of the standard Dale and Reiter generation algorithm for definite descriptions: one to deal with inference and the generation of bridging descriptions, the other to address the problems raised by incrementality. The paper concludes by discussing the results of a corpus study which gives some pointers for further research on the generation of referring expressions.
Keywords
referring expression generation algorithms
inference and natural language processing
constraint-based language processing
corpus study
inférence et traitement automatique des langues
approches basées sur les contraintes
descriptions définies
analyse de corpus
definite descriptions
génération d'expressions référentielles