Integrated natural language generation with schema-tree adjoining grammars

Abstract

This paper describes an integrated generation system (INLGS) based on the formalism of Schema Tree Adjoining Grammars with Unification (SU-TAGs). According to this system architecture, all knowledge bases are specified in the same formalism and run the same processing algorithm. A main advantage is that negotiation between generation components can easily be imposed on the system. Moreover, only one algorithm must be implemented and tested in order to provide the one and only processing unit. In the INLGS a reversible parser/generator is deployed. It runs knowledge bases in the formalism of SU-TAGs. SU- TAG comprises a condensed grammar representation and direct parsing/ generation deals with partially unspecified schemata. Instead of developing new knowledge bases from scratch, existing ones are reused here. This means all knowledge bases of the INLGS are transformed (e.g., the TAG-based XTAG system and the plan-based interpersonal model of VOTE).