Enhancing Empirical Research for Linguistically Motivated Precision Grammars

Grammars of natural language are highly complex objects. This complexity is reflected in formal analyses found in both syntactic theory and computational grammars. In particular, there are two factors that make it notoriously difficult to make strong assertions about analyses for natural language grammars. First, syntactic phenomena interact and therefore also their analyses interact. Second, typically more than one formal analysis can account for a given phenomenon. Grammar engineering is an incremental process. A syntactician or grammar engineer often needs to choose between analyses without conclusive evidence as to which analysis works best. Because analyses interact, this decision will affect possible analyses for phenomena handled in the future. This thesis proposes to store analyses in a metagrammar allowing grammar writers to maintain alternative analyses until enough evidence is found to make a choice. The approach is evaluated by comparing alternative analyses for auxiliaries and word order in German. Additionally, possibilities of using the approach for multilingual grammar engineering are examined. The results of these evaluations indicate that the proposed methodology provides a platform for a more systematic approach for grammar engineering. The approach furthermore improves modularity offering possibilities for parallel grammar development for multiple (related) languages.