Generating Swiss German sentences from Standard German: a multi-dialectal approach

Abstract

This thesis proposes to combine methods and data from two rather distant fields of language science – dialectology and human language technology – into a system that automatically transforms Standard German words and sentences into multiple Swiss German dialects. Our work is inspired by previous research in generative dialectology and computational linguistics, which attempts to derive multiple dialect systems from a single reference system with the help of hand-written transformation rules. We propose to call such rules 'georeferenced', in the sense that they are linked to probability maps that specify their area of validity. These probability maps are extracted by interpolation from existing dialectological atlases. Finally, as a consequence of our map digitization efforts, we are able to present original dialectometrical results for the Swiss German dialect landscape.