Entailment Relation Aware Paraphrase Generation

TL;DR

This paper presents ERAP, a reinforcement learning-based system for entailment relation aware paraphrase generation that improves data augmentation for textual entailment tasks by explicitly controlling the entailment relation.

Contribution

It introduces a novel task of entailment relation aware paraphrasing and proposes ERAP, a weakly-supervised system trained without task-specific data, to generate high-quality, relation-specific paraphrases.

Findings

ERAP generates paraphrases conforming to specified entailment relations.

Using ERAP for data augmentation improves textual entailment performance.

ERAP produces fewer artifacts compared to uncontrolled paraphrasing systems.

Abstract

We introduce a new task of entailment relation aware paraphrase generation which aims at generating a paraphrase conforming to a given entailment relation (e.g. equivalent, forward entailing, or reverse entailing) with respect to a given input. We propose a reinforcement learning-based weakly-supervised paraphrasing system, ERAP, that can be trained using existing paraphrase and natural language inference (NLI) corpora without an explicit task-specific corpus. A combination of automated and human evaluations show that ERAP generates paraphrases conforming to the specified entailment relation and are of good quality as compared to the baselines and uncontrolled paraphrasing systems. Using ERAP for augmenting training data for downstream textual entailment task improves performance over an uncontrolled paraphrasing system, and introduces fewer training artifacts, indicating the benefit of explicit control during paraphrasing.