Querying topological fields in the TIGER scheme with TIGERSearch

The TIGER corpus is a German treebank with a hybrid dependency-constituency annotation. In this paper, I address the question how well topological fields (e.g. Vorfeld, Verb second) can be searched in this treebank, using the search tool TIGERSearch. For most queries, a version without crossing branches is used. It turns out that queries can be formulated that result in quite good F-scores for the Vorfeld and left and right brackets. Mittelfeld and Nachfeld are hard to query. This is due partly to properties of the language, partly to design decisions in the TIGER scheme, and partly to restrictions imposed by the search tool.