De-Constraining Text Generation
Stephen Beale, Sergei Nirenburg, Evelyne Viegast, Leo Wanner · 1998
We argue that the current, predominantly task-oriented, approaches to modularizing text generation, while plausible and useful conceptually, set up spurious conceptual and operational constraints. We propose a data-driven approach to modularization and illustrate how it eliminates the previously ubiquitous constraints on combination of evidence across modules and on control. We also briefly overview the constraint-based control architecture that enables such an approach and facilitates near linear-time processing with realistic texts. 1 Introduction This paper addresses the area of text generation known as microplanning [Levelt1989, Panaget1994, Huang and Fiedler1996], or sentence planning [Rambow and Korelsky1992]; [Wanner and Hovy1996]. Microplanning involves low-level discourse structuring and marking, sentence boundary planning, clause internal structuring and all of the varied subtasks involved in lexical choice. Conventional wisdom dictates that these complex tasks be modularize...