A Role and Reference Grammar parser for German

Abstract

This paper describes the development of a sentence parser for German based on Role and Reference Grammar (RRG). German has a number of characteristics which are very challenging to a computational approach to syntactic processing that is semantically motivated. These include the German sentence structure, which is organised as a “bracket structure”, in which one component of the predicate, mostly the finite verb, is realised in an early position in the sentence and therefore forms the “left bracket”, whereas the more meaningful part of the predicate, mostly the infinite predicate, appears at the end of the sentence and forms the “right bracket”. Inside these brackets, the main information of the sentence is placed in an order that reflects information structure, i.e. the given information precedes the new information inside the bracket. The bracket structure has recurrent formal elements. It therefore provides a formal structuring principle for sentences that is particular for German and serves functions for sentence processing, information structure and turn taking. German has a three-way gender system for nouns. Gender is marked on the article. Tense is marked on the verb. There is an aspectual system in the realisation of perfect in that telic verbs select sein in the perfect, while atelic verbs select haben. The software is developed using Java and will accept an input sentence in German from the user through a Java Swing graphical user interface. The software parses an input sentence into its respective tokens using the Java native string processing capability. Each token is checked against the respective German lexicons (noun, verb, etc.) to determine its lexical category. The morphological markings of the tokens are unpacked for their feature sets. These features are recorded for each token. The lexical entry for the verb is determined and its logical structure is retrieved from the lexicon. The result of the parse displays the populated RRG logical structure and its operators, i.e. tense, aspect etc. A trace of the application of the linking system, in mapping from syntactic clause to semantic representation, is presented.