Everything Happens for a Reason: Discovering the Purpose of Actions in Procedural Text

TL;DR

This paper introduces XPAD, a model that improves understanding of procedural texts by predicting action effects and their dependencies, explaining the reasoning behind actions, and extending a benchmark dataset for better evaluation.

Contribution

The paper presents XPAD, a novel model that predicts action effects and dependencies, and extends the ProPara dataset with new explanation tasks for procedural text comprehension.

Findings

XPAD outperforms previous systems on action dependency prediction.

XPAD maintains performance on original procedural comprehension tasks.

Extended ProPara dataset enables better evaluation of action explanations.

Abstract

Our goal is to better comprehend procedural text, e.g., a paragraph about photosynthesis, by not only predicting what happens, but why some actions need to happen before others. Our approach builds on a prior process comprehension framework for predicting actions' effects, to also identify subsequent steps that those effects enable. We present our new model (XPAD) that biases effect predictions towards those that (1) explain more of the actions in the paragraph and (2) are more plausible with respect to background knowledge. We also extend an existing benchmark dataset for procedural text comprehension, ProPara, by adding the new task of explaining actions by predicting their dependencies. We find that XPAD significantly outperforms prior systems on this task, while maintaining the performance on the original task in ProPara. The dataset is available at http://data.allenai.org/propara