Visual Information Guided Zero-Shot Paraphrase Generation

TL;DR

This paper introduces ViPG, a novel zero-shot paraphrase generation method that uses images as pivots, jointly training image captioning and paraphrasing models to improve relevancy, fluency, and diversity.

Contribution

It proposes a new visual information guided approach for zero-shot paraphrase generation using paired image-caption data, differing from traditional pipeline methods.

Findings

Model achieves high relevancy, fluency, and diversity in generated paraphrases.

Both automatic and human evaluations validate the effectiveness of using images as pivots.

Image-based pivoting is a promising direction for zero-shot paraphrase generation.

Abstract

Zero-shot paraphrase generation has drawn much attention as the large-scale high-quality paraphrase corpus is limited. Back-translation, also known as the pivot-based method, is typical to this end. Several works leverage different information as "pivot" such as language, semantic representation and so on. In this paper, we explore using visual information such as image as the "pivot" of back-translation. Different with the pipeline back-translation method, we propose visual information guided zero-shot paraphrase generation (ViPG) based only on paired image-caption data. It jointly trains an image captioning model and a paraphrasing model and leverage the image captioning model to guide the training of the paraphrasing model. Both automatic evaluation and human evaluation show our model can generate paraphrase with good relevancy, fluency and diversity, and image is a promising kind of pivot for zero-shot paraphrase generation.