Bridging Knowledge Gaps in Neural Entailment via Symbolic Models

TL;DR

This paper introduces NSnet, a neural entailment model that integrates structured scientific knowledge to better identify entailment, addressing knowledge gaps often overlooked by traditional models.

Contribution

The paper presents a novel architecture combining neural models with a knowledge lookup module for scientific entailment tasks, improving performance by leveraging external structured knowledge.

Findings

NSnet outperforms base models by 5% on SciTail.

Knowledge integration improves entailment accuracy.

Fact-level decomposition aids in verifying hypotheses.

Abstract

Most textual entailment models focus on lexical gaps between the premise text and the hypothesis, but rarely on knowledge gaps. We focus on filling these knowledge gaps in the Science Entailment task, by leveraging an external structured knowledge base (KB) of science facts. Our new architecture combines standard neural entailment models with a knowledge lookup module. To facilitate this lookup, we propose a fact-level decomposition of the hypothesis, and verifying the resulting sub-facts against both the textual premise and the structured KB. Our model, NSnet, learns to aggregate predictions from these heterogeneous data formats. On the SciTail dataset, NSnet outperforms a simpler combination of the two predictions by 3% and the base entailment model by 5%.