R$^2$F: A General Retrieval, Reading and Fusion Framework for Document-level Natural Language Inference

TL;DR

This paper introduces R2F, a framework that simplifies document-level natural language inference by breaking it into sentence-level tasks, improving interpretability and performance on longer documents.

Contribution

The paper proposes a novel Retrieval, Reading and Fusion (R2F) framework and a new setting for DOCNLI, addressing interpretability, long-range dependency, and cross-sentence inference challenges.

Findings

Achieves state-of-the-art performance on DOCNLI tasks.

Enhances interpretability with evidence-based reasoning.

Robust across diverse evidence retrieval methods.

Abstract

Document-level natural language inference (DOCNLI) is a new challenging task in natural language processing, aiming at judging the entailment relationship between a pair of hypothesis and premise documents. Current datasets and baselines largely follow sentence-level settings, but fail to address the issues raised by longer documents. In this paper, we establish a general solution, named Retrieval, Reading and Fusion (R2F) framework, and a new setting, by analyzing the main challenges of DOCNLI: interpretability, long-range dependency, and cross-sentence inference. The basic idea of the framework is to simplify document-level task into a set of sentence-level tasks, and improve both performance and interpretability with the power of evidence. For each hypothesis sentence, the framework retrieves evidence sentences from the premise, and reads to estimate its credibility. Then the sentence-level results are fused to judge the relationship between the documents. For the setting, we contribute complementary evidence and entailment label annotation on hypothesis sentences, for interpretability study. Our experimental results show that R2F framework can obtain state-of-the-art performance and is robust for diverse evidence retrieval methods. Moreover, it can give more interpretable prediction results. Our model and code are released at https://github.com/phoenixsecularbird/R2F.