Discourse Framework

Discourse frameworks analyze how sentences relate within a larger text or conversation, aiming to understand the underlying structure and meaning beyond individual sentences. Current research focuses on developing computational models, often employing machine learning techniques like joint training and contrastive learning, to automatically identify and categorize discourse elements such as questions under discussion, stance-taking, and counterspeech. These advancements have implications for various applications, including automated writing evaluation, hate speech detection, and improving the coherence of automatically generated text in diverse contexts like news articles and video captions.