Discourse Relation Prediction and Discourse Parsing in Dialogues with Minimal Supervision - Archive ouverte HAL Access content directly
Conference Papers Year : 2024

Discourse Relation Prediction and Discourse Parsing in Dialogues with Minimal Supervision

Abstract

Discourse analysis plays a crucial role in Nat- ural Language Processing, with discourse re- lation prediction arguably being the most dif- ficult task in discourse parsing. Previous stud- ies have generally focused on explicit or im- plicit discourse relation classification in mono- logues, leaving dialogue an under-explored do- main. Facing the data scarcity issue, we pro- pose to leverage self-training strategies based on a Transformer backbone. Moreover, we design the first semi-supervised pipeline that sequentially predicts discourse structures and relations. Using 50 examples, our relation pre- diction module achieves 58.4 in accuracy on the STAC corpus, close to supervised state-of- the-art. Full parsing results show notable im- provements compared to the supervised mod- els both in-domain (gaming) and cross-domain (technical chat), with better stability.
Fichier principal
Vignette du fichier
Lisa_eacl2024__codi24.pdf (763.66 Ko) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-04524155 , version 1 (27-03-2024)

Identifiers

  • HAL Id : hal-04524155 , version 1

Cite

Chuyuan Li, Chloé Braud, Maxime Amblard, Giuseppe Carenini. Discourse Relation Prediction and Discourse Parsing in Dialogues with Minimal Supervision. 5th Workshop on Computational Approaches to Discourse (CODI 2024) co-located with EACL 2024, Heidelberg Institute for Theoretical Studies (HITS), Mar 2024, Malte, Malta. ⟨hal-04524155⟩
13 View
10 Download

Share

Gmail Facebook X LinkedIn More