When natural language processing stretches beyond the sentence level, we are entering the realm of discourse: In dialogue, speakers take turns and make contributions to the unfolding interaction, which can be modeled as a sequence of moves. In monologue, the individual sentences make contributions to the unfolding text meaning, involving notions such as coreference and coherence relations (amongst others).
Discourse processing thus refers to those tasks of natural language processing that operate across sentence boundaries: resolving anaphors, determining topic boundaries, identifying coherence relations, and so on.
In the Applied CompLing Discourse Research Lab, directed by Manfred Stede, we undertake research on both theoretical and applied aspects of discourse processing: On the theoretical side, our work revolves around modeling text structure as a multi-layer phenomenon, with a current focus on coherence relations and their linguistic signals. On the applied side, we work on tasks like text summarization, classification, genre-sensitive information extraction, sentiment analysis, and text generation. Finally, we have been active in annotation research and tool development for many years.