Abstract
The paper reports on a developing project concerning manual annotation of discourse relations for Czech. The aim of the project is to design a language corpus capturing Czech language material from the perspective of text structure and coherence, i.e. focusing on the description of inter-sententional relations.
The outcome of the project will be a new annotation layer above the existing layers of annotation (morphology, surface syntax and underlying syntax) in the Prague Dependency Treebank. This discourse annotation should function as a unique source for linguistic research in the field of the discourse analysis and for computational experiments in text processing and summarization as well as machine translation.