Needles in Haystacks: Semi-Automatic Identification of Regional Grammatical Variation in Standard German

This paper lays out a semi-automatic approach to identifying regional variation in the grammar of Standard German. Our approach takes as input manually defined templates of grammatical constructions that are auto- matically instantiated over a corpus collected from regional newspapers. These instantiations are automatically ranked by a metric that quantifies how spe- cific an instantiation is for a region. Ranked lists of instantiations are compiled that contain instantiations specific to a region and are scanned manually by linguists to identify those that denote grammatical variants of Standard Ger- man. This approach enabled us to discover variants that so far have not been documented. With respect to research on variation within standard languages as seen from a more general perspective, we aim to contribute towards research strategies that clearly rely on empiricism rather than on intuition or bias