Constraint-based Text Generation Using Local Constraints and Argumentation to Generate a Turn in Conversation

Elhadad, Michael

In this paper, I present a natural language generation system aimed at constructing a turn within an ongoing conversation. Three main points are advanced: (1) previous discourse determines the production of a new turn; this paper describes techniques to take the influence of previous discourse into account when generating text. (2) the connection between previous discourse and a new turn is described as the interaction between five local constraints; local constraints are viewed as relations between one discourse segment of the previous discourse and the new turn. (3) argumentation is one class of local constraints which has important effects on the form of the language produced and interacts closely with the other types of local constraints used in this work. The approach is applied to the implementation of an explanation module for the ADVISOR expert system.



More About This Work

Academic Units
Computer Science
Department of Computer Science, Columbia University
Columbia University Computer Science Technical Reports, CUCS-003-90
Published Here
March 2, 2012