Principled Paraphrase Generation with Parallel Corpora

TL;DR

This paper introduces a novel, principled method for paraphrase generation that improves upon round-trip machine translation by using an information bottleneck approach, resulting in better quality and control over paraphrase diversity.

Contribution

It proposes an alternative similarity metric and a new training method with an adversarial term, enabling more effective and efficient paraphrase generation without pivot translations.

Findings

Outperforms round-trip MT in quality and efficiency

Provides adjustable control over fidelity-diversity trade-off

Achieves better results in experiments

Abstract

Round-trip Machine Translation (MT) is a popular choice for paraphrase generation, which leverages readily available parallel corpora for supervision. In this paper, we formalize the implicit similarity function induced by this approach, and show that it is susceptible to non-paraphrase pairs sharing a single ambiguous translation. Based on these insights, we design an alternative similarity metric that mitigates this issue by requiring the entire translation distribution to match, and implement a relaxation of it through the Information Bottleneck method. Our approach incorporates an adversarial term into MT training in order to learn representations that encode as much information about the reference translation as possible, while keeping as little information about the input as possible. Paraphrases can be generated by decoding back to the source from this representation, without having to generate pivot translations. In addition to being more principled and efficient than round-trip MT, our approach offers an adjustable parameter to control the fidelity-diversity trade-off, and obtains better results in our experiments.