Submission 460
Narrative Coherence Through Multiple Perspectives
SymposiumTalk-03
Presented by: Frances Yung
Understanding how readers construct meaning from text requires accounting for the coherence relations that link discourse segments. Yet the interpretation of these relations is shaped by the reader’s own perspective. In this talk, I present perspectivist analyses of coherence interpretation based on two complementary methodologies: linguistic annotation and computational modeling. Drawing on annotation studies across translations, I show how translational choices introduce additional layers of interpretive nuance and highlight cross-linguistic variation in coherence perception. I then turn to computational models designed to capture individual perspectives on discourse relation recognition, leveraging multi-annotator data collected through crowdsourcing. Together, these findings reveal the variability inherent in discourse coherence and point toward individuality-aware reader models that can more faithfully represent narrative processing across domains.