Feature Rich Models for Discourse Signaling - 2018

Explore feature-rich models for discourse signaling in this comprehensive lecture by Amir Zeldes from Georgetown University. Delve into the complexities of discourse relations such as contrast, cause, and evidence, examining how humans understand sentence functions in relation to each other. Investigate the challenges of recognizing discourse relations, including ambiguous or implicit signals, and learn about computational models designed to account for these complexities. Discover the limitations of purely text-based models using RNNs and word embeddings, and understand the importance of incorporating richly annotated data beyond the textual level. Gain insights into the Rhetorical Structure Theory framework and explore data from the RST Signaling Corpus and the GUM corpus. This talk emphasizes the need for syntactic and semantic information to form a more complete picture of discourse relations in text, providing valuable knowledge for researchers and practitioners in computational linguistics and discourse analysis.