Skip to main content
Paper

A Cross-Genre Analysis of Discourse Relation Signaling in the GUM Corpus

Author
  • Lauren Levine (Georgetown University)

Abstract

In this paper, we investigate the cross-genre variation in how discourse relations are signaled in the Georgetown University Multilayer (GUM) Corpus, an English language corpus which contains 16 different genres of texts with various linguistic annotations, including Rhetorical Structure Theory (RST) style discourse annotations. We look at the proportions of discourse signals in each genre, and then we conduct an analysis of which discourse relations display the most inter-genre variation in how they are signaled, providing a methodology for ranking the inter-genre variability of the signaling of individual discourse relations. Although the way in which individual discourse relations are signaled in GUM is relatively stable across genres, we are able still to produce stable rankings, finding that organizationrestatement, and explanation relations display the most inter-genre variation. However, we find that genre specific graphical norms can account for a large portion of the observed variation. 

Keywords: RST, signaling, discourse relations, genre, variation, eRST, discourse relation signaling, corpus analysis, ranking variation

How to Cite:

Levine, L., (2025) “A Cross-Genre Analysis of Discourse Relation Signaling in the GUM Corpus”, Society for Computation in Linguistics 8(1): 30. doi: https://doi.org/10.7275/scil.3155

Downloads:
Download PDF

49 Views

7 Downloads

Published on
2025-06-12

Peer Reviewed