Julia Koch, Toni Bernhart, André Blessing, Gunilla Eschenbach, Markus Gärtner, Kerstin Jung, Nora Ketschik, Anna Kinder, Jonas Kuhn, Sandra Richter, Nadja Schauffler, Rebecca Sturm, Gabriel Viehhauser, Ngok Thang Vu
doi.org/10.7592/PP2025.06.koch_et_al
At the heart of the »textklang« (Sound of Text) project is the development of a mixed-method approach to investigate the interrelation between written lyric poetry and its sonic realisation. It is an interdisciplinary collaboration between the German Literary Archive in Marbach and the University of Stuttgart that includes literary studies, digital humanities, computational linguistics, laboratory phonology and speech technology. The project’s corpus is centred on the poetry of Romanticism and is based on the holdings of the German Literature Archive. In our contribution, we illustrate a multi-perspective approach to one collection within the corpus entitled The Boy’s Magic Horn (Des Knaben Wunderhorn), edited in three volumes by A. v. Arnim and C. Brentano in 1806 and 1808 and including more than 700 poems. The Boy’s Magic Horn is considered one of the most influential poetry collections in German literature because of its vivid reception in both ‘low’ folkloristic cultures and ‘high’ culture, especially in musical settings (G. Mahler, J. Brahms, F. Silcher). We share some initial outcomes of the ongoing research process considering quantitative, textual, prosodic, and sonic aspects. As part of the project’s methodological and experimental toolbox, we present a speech synthesis model that has been trained on this sub-corpus and which results in a better realisation of poetic speech compared to synthesis models exclusively trained on prose data. Finally, we discuss the challenges this data pose to automatic processing tools.
