Diskursrepräsentationen

Diskursrepräsentationen (discourse representations) are theoretical constructs used in discourse analysis and computational linguistics to model the meaning of discourse. They aim to capture how the meaning of sentences and larger texts is built up and maintained through the interplay of linguistic elements and their context. The central idea is that discourse is not simply a sum of individual sentence meanings, but rather a dynamic process where meaning is constructed incrementally.

Different approaches exist to formalize discourse representations. One prominent framework is Discourse Representation Theory (DRT), developed by Hans Kamp and Irene Heiner-Möller. DRT uses a formal language with variables and discourse referents to represent the information conveyed by a text. These representations are typically structured in a way that reflects the accessibility and anaphora resolution of entities mentioned in the discourse. For instance, a pronoun's reference is resolved by linking it to a previously introduced discourse referent.

Beyond DRT, other methods and frameworks also focus on representing discourse meaning. These can involve semantic networks, frame-based approaches, or more recent neural network-based representations that capture contextualized word meanings. The goal across these approaches is to provide a systematic way to understand how discourse coherence is achieved, how information is updated, and how the overall meaning of a text emerges from the combination of its parts. Diskursrepräsentationen are thus crucial for tasks such as question answering, summarization, and machine translation, where a deep understanding of discourse is essential.