Morphological Analysis in Integrated Natural Language Interfaces to Deductive Databases

The aim of our research is the implementation of a prototype for a German natural language interface to a deductive database system modelling a production planning and control system. Whereas existent natural language interfaces represent only loosely coupled filters separated from the underlying database, in our approach the interface constitutes an integral part of the database system itself by making use of the declarative power of logic programming provided by deductive databases. The paper focuses on the already completed morphological analysis. As a consequence of our design we adapted the lexical approach by storing only canonical forms as dictionary entries and attaching to them all morphological features not only considering word endings but also covering the complex cases of prefixes, derivations, and compound words. Furthermore, the expressive two-level formalism was applied to the processing of some special morphological phenomena (ablaut, elision, binding sounds). Our morphological analysis turned out to be a powerful and easily extendible tool, constituting an excellent basis for further syntactic and semantic analysis. It provides the possibility of assigning semantic features to morphological units at the appropriate level of the hierarchically structured dictionary by making use of inheritance in order to achieve a natural and concise representation of meaning.