Abstract

Short epics in Middle High German are a challenge for author attribution procedures as they involve dealing with short texts, incomplete normalisation, and generic tests. Firstly, the paper discusses optimisation methods using basic form and grammar tags extracted from the ‘Mittelhochdeutsche Begriffsdatenbank’ (Middle High German Conceptual Database). As part of this, an evaluation test with normalised Middle High German texts is carried out where 20 texts with known authorship are used as ‘guessing texts’ and a further 19 works of the same authors as ‘comparison texts’. The use of grammar tags or basic forms, instead of inflected word forms leads to a deterioration of the recognition rate and the combination of basic forms or grammar tags with word forms leads to a slight improvement. Secondly, the authorship of the ‘Halbe Birne’ will be investigated with eight texts by Konrad von Würzburg, as well as a comparative corpus of 14 tales (German ‘Mären’) and seven novels. This leads to a direct calculation of which words can be specifically ascribed to Konrad von Würzburg. This method is first evaluated by determining the distinctive words on four Konrad texts; a control test on four other Konrad texts shows that a preferred use of the distinctive words leads to a significant increase in the recognition rate. Then, a Konrad-specific vocabulary is determined on the basis of eight Konrad texts. Using these distinctive words it can finally be shown that Konrad can be attributed with some certainty as the author of the ‘Halbe Birne’.