notesum.ai

Published at November 22

Leveraging Hierarchical Prototypes as the Verbalizer for Implicit Discourse Relation Recognition

cs.CL

Released Date: November 22, 2024

Authors: Wanqiu Long, Bonnie Webber¹

Aff.: ¹University of Edinburgh, Edinburgh, UK

Arxiv: http://arxiv.org/abs/2411.14880v1

Refer to caption

Model	PDTB-2
	Top Level		Second Level
	Acc	Macro-F1	Acc	Macro-F1
Dai and Huang (2019)	59.66	52.89	48.23	33.41
Shi and Demberg (2019a)	61.42	46.40	47.83	-
Nguyen et al. (2019)	-	53.00	49.95	-
Guo et al. (2020)	57.25	47.90	-	-
Kishimoto et al. (2020)	65.26	58.48	52.34	-
Liu et al. (2020)	69.06	63.39	58.13	-
Jiang et al. (2021)	-	57.18	-	37.76
Dou et al. (2021)	70.17	65.06	-	-
Wu et al. (2022)	71.18	63.73	60.33	40.49
Long and Webber (2022)	71.70	67.85	59.19	45.54
Zhou et al. (2022)	70.84	64.95	60.54	41.55
Chan et al. (2023)(T5-base)	71.70	65.79	61.02	43.68
Ours (RoBERTa-base)	72.47	69.66	60.73	47.07