Relational Gating for "What If" Reasoning

TL;DR

This paper introduces a relational gating network that improves procedural reasoning over text for answering 'What if...' questions by filtering key entities and relationships, capturing higher-order relations, and achieving state-of-the-art results.

Contribution

The paper presents a novel relational gating network with entity, relation, and contextual modules for enhanced procedural reasoning in text understanding.

Findings

Achieves state-of-the-art results on WIQA dataset.

Model effectively captures higher-order relations and causal reasoning.

Outperforms previous methods in 'What if...' question answering.

Abstract

This paper addresses the challenge of learning to do procedural reasoning over text to answer "What if..." questions. We propose a novel relational gating network that learns to filter the key entities and relationships and learns contextual and cross representations of both procedure and question for finding the answer. Our relational gating network contains an entity gating module, relation gating module, and contextual interaction module. These modules help in solving the "What if..." reasoning problem. We show that modeling pairwise relationships helps to capture higher-order relations and find the line of reasoning for causes and effects in the procedural descriptions. Our proposed approach achieves the state-of-the-art results on the WIQA dataset.