Discourse-Aware Graph Networks for Textual Logical Reasoning



Huang, Yinya, Liu, Lemao, Xu, Kun, Fang, Meng ORCID: 0000-0001-6745-286X, Lin, Liang and Liang, Xiaodan
(2023) Discourse-Aware Graph Networks for Textual Logical Reasoning. IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 45 (10). pp. 11668-11688.

[thumbnail of 2207.01450.pdf] Text
2207.01450.pdf - Author Accepted Manuscript

Download (18MB) | Preview

Abstract

Textual logical reasoning, especially question-answering (QA) tasks with logical reasoning, requires awareness of particular logical structures. The passage-level logical relations represent entailment or contradiction between propositional units (e.g., a concluding sentence). However, such structures are unexplored as current QA systems focus on entity-based relations. In this work, we propose logic structural-constraint modeling to solve the logical reasoning QA and introduce discourse-aware graph networks (DAGNs). The networks first construct logic graphs leveraging in-line discourse connectives and generic logic theories, then learn logic representations by end-to-end evolving the logic relations with an edge-reasoning mechanism and updating the graph features. This pipeline is applied to a general encoder, whose fundamental features are joined with the high-level logic features for answer prediction. Experiments on three textual logical reasoning datasets demonstrate the reasonability of the logical structures built in DAGNs and the effectiveness of the learned logic features. Moreover, zero-shot transfer results show the features' generality to unseen logical texts.

Item Type: Article
Uncontrolled Keywords: Natural language processing, logical reasoning, question answering, multi-turn dialogue reasoning, graph neural networks, supervised learning, zero-shot learning
Divisions: Faculty of Science and Engineering > School of Electrical Engineering, Electronics and Computer Science
Depositing User: Symplectic Admin
Date Deposited: 04 Oct 2023 08:22
Last Modified: 31 Oct 2023 02:10
DOI: 10.1109/TPAMI.2023.3280178
Related URLs:
URI: https://livrepository.liverpool.ac.uk/id/eprint/3173362