Interfacing Constraint-Based Grammars and Generation Algorithms

TL;DR

This paper addresses the challenge of interfacing constraint-based grammars with generation algorithms by proposing a control-oriented layer to improve semantic organization for generation.

Contribution

It introduces a novel approach that explicitly manages the grammar-generation interface, enhancing semantic compatibility for surface generation tasks.

Findings

Semantic analyses can hinder generation algorithms like SHDG.

Adding a control layer improves semantic reorganization for generation.

The approach is demonstrated using DISCO German grammar.

Abstract

Constraint-based grammars can, in principle, serve as the major linguistic knowledge source for both parsing and generation. Surface generation starts from input semantics representations that may vary across grammars. For many declarative grammars, the concept of derivation implicitly built in is that of parsing. They may thus not be interpretable by a generation algorithm. We show that linguistically plausible semantic analyses can cause severe problems for semantic-head-driven approaches for generation (SHDG). We use SeReal, a variant of SHDG and the DISCO grammar of German as our source of examples. We propose a new, general approach that explicitly accounts for the interface between the grammar and the generation algorithm by adding a control-oriented layer to the linguistic knowledge base that reorganizes the semantics in a way suitable for generation.