A Contrastive Evaluation of Functional Unification Grammar for Surface Language Generation: A Case Study in Choice of Connectives

Abstract

Language generation systems have used a variety of grammatical formalisms for producing syntactic structure and yet, there has been little research evaluating the formalisms for the specifics of the generation task. In our work at Columbia we have primarily used a unification based formalism, a functional Unification Grammar (FUG) and have found it well suited for many of the generation tasks we have addressed. FUG turned out to be better suited for generation than various off-the-shelf parsing formalisms (including an Augmented Transition Network, a Bottom-Up Chart Parser, and a declarative Clause Grammar) that we have also used over the past 5 years. In this paper, we identify the characteristics of FUG that we find useful for generation and focus on order of decision making and its impact on expressions of constraints. Our claim is that order of decision making in FUG through unification allows for flexible interaction between constraints, which, in turn, allows for a more concise representation of constraints. To illustrate these properties of FUG, we use a subtask of the lexical choice problem, namely the task of selecting a connective (e.g., but, however, nonetheless, since, because, etc.) to conjoin two input propositions.