Dear Sir or Madam, May I introduce the GYAFC Dataset: Corpus, Benchmarks and Metrics for Formality Style Transfer

TL;DR

This paper introduces the GYAFC dataset for formality style transfer, providing a large corpus, benchmarks, and discussing challenges with automatic metrics to advance research in style transfer.

Contribution

It presents the largest corpus for formality style transfer, adapts machine translation techniques as baselines, and analyzes issues with automatic evaluation metrics.

Findings

Established strong baseline methods from machine translation

Created the largest formality transfer dataset to date

Highlighted challenges in automatic metric evaluation

Abstract

Style transfer is the task of automatically transforming a piece of text in one particular style into another. A major barrier to progress in this field has been a lack of training and evaluation datasets, as well as benchmarks and automatic metrics. In this work, we create the largest corpus for a particular stylistic transfer (formality) and show that techniques from the machine translation community can serve as strong baselines for future work. We also discuss challenges of using automatic metrics.