The segmentation of texts into Elementary Discourse Units (EDUs) is being tested in the COMUTE project as a preparatory step for alignment. In their journal article ‘Discourse Segmentation of German Text with Pretrained Language Models’, Steffen Frenzel and Manfred Stede present the results of a manual study and experiments with various segmentation models on the text data from the research project.

The paper can be accessed here: https://jlcl.org/article/view/306