Incremental Syntactic Generation of Natural Language with Tree Adjoining Grammars

Anne Schauder

DFKI DFKI Documents (D) 92-21 1992.


This document combines the basic ideas of my master's thesis---which has been developped within the WIP project---with new results from my work as a member of WIP, as far as they concern the integration and further development of the implemented system. ISGT (in German 'Inkrementeller Syntaktischer Generierer natürlicher Sprache mit TAGs') is a syntactic component for a text generation system and is based on Tree Adjoining Grammars. It is lexically guided and consists of two levels of syntactic processing: A component that computes the hierarchical structure of the sentence under construction (hierarchical level) and a component that computes the word position and utters the sentence (positional level). The central aim of this work has been to design a syntactic generator that computes sentences in an incremental fashion. The realization of the incremental syntactic generator has been supported by a distributed parallel model that is used to speed up the computation of single parts of the sentence.

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence