Improving Paraphrase Detection with the Adversarial Paraphrasing Task

TL;DR

This paper introduces an adversarial paraphrasing task and dataset to improve paraphrase detection models by emphasizing semantic equivalence over lexical or syntactic similarity, leading to better understanding of sentence meaning.

Contribution

The paper proposes a novel adversarial dataset creation method for paraphrase detection that focuses on semantic equivalence, and demonstrates its effectiveness in enhancing model performance.

Findings

Adversarial paraphrase dataset improves model accuracy.

Automated dataset generation using T5 accelerates data creation.

Models trained on the dataset better detect sentence-level meaning equivalence.

Abstract

If two sentences have the same meaning, it should follow that they are equivalent in their inferential properties, i.e., each sentence should textually entail the other. However, many paraphrase datasets currently in widespread use rely on a sense of paraphrase based on word overlap and syntax. Can we teach them instead to identify paraphrases in a way that draws on the inferential properties of the sentences, and is not over-reliant on lexical and syntactic similarities of a sentence pair? We apply the adversarial paradigm to this question, and introduce a new adversarial method of dataset creation for paraphrase identification: the Adversarial Paraphrasing Task (APT), which asks participants to generate semantically equivalent (in the sense of mutually implicative) but lexically and syntactically disparate paraphrases. These sentence pairs can then be used both to test paraphrase identification models (which get barely random accuracy) and then improve their performance. To accelerate dataset generation, we explore automation of APT using T5, and show that the resulting dataset also improves accuracy. We discuss implications for paraphrase detection and release our dataset in the hope of making paraphrase detection models better able to detect sentence-level meaning equivalence.