A Multi-Task and Multi-Label Classification Model for Implicit Discourse Relation Recognition

TL;DR

This paper introduces a multi-task, multi-label classification model for implicit discourse relation recognition, achieving state-of-the-art results and establishing a new benchmark in the field.

Contribution

It presents the first multi-label IDRR model, adaptable to single-label tasks, with extensive experiments and transfer learning analysis between datasets.

Findings

Achieved SOTA results on single-label IDRR with DiscoGeM.

Established the first benchmark for multi-label IDRR.

Provided analysis of transfer learning between DiscoGeM and PDTB 3.0.

Abstract

We propose a novel multi-label classification approach to implicit discourse relation recognition (IDRR). Our approach features a multi-task model that jointly learns multi-label representations of implicit discourse relations across all three sense levels in the PDTB 3.0 framework. The model can also be adapted to the traditional single-label IDRR setting by selecting the sense with the highest probability in the multi-label representation. We conduct extensive experiments to identify optimal model configurations and loss functions in both settings. Our approach establishes the first benchmark for multi-label IDRR and achieves SOTA results on single-label IDRR using DiscoGeM. Finally, we evaluate our model on the PDTB 3.0 corpus in the single-label setting, presenting the first analysis of transfer learning between the DiscoGeM and PDTB 3.0 corpora for IDRR.