Exploiting Social Media Content for Self-Supervised Style Transfer

Dana Ruiter, Thomas Kleinbauer, Cristina España-Bonet, Josef van Genabith, Dietrich Klakow


Abstract
Recent research on style transfer takes inspiration from unsupervised neural machine translation (UNMT), learning from large amounts of non-parallel data by exploiting cycle consistency loss, back-translation, and denoising autoencoders. By contrast, the use of selfsupervised NMT (SSNMT), which leverages (near) parallel instances hidden in non-parallel data more efficiently than UNMT, has not yet been explored for style transfer. In this paper we present a novel Self-Supervised Style Transfer (3ST) model, which augments SSNMT with UNMT methods in order to identify and efficiently exploit supervisory signals in non-parallel social media posts. We compare 3ST with state-of-the-art (SOTA) style transfer models across civil rephrasing, formality and polarity tasks. We show that 3ST is able to balance the three major objectives (fluency, content preservation, attribute transfer accuracy) the best, outperforming SOTA models on averaged performance across their tested tasks in automatic and human evaluation.
Anthology ID:
2022.socialnlp-1.2
Volume:
Proceedings of the Tenth International Workshop on Natural Language Processing for Social Media
Month:
July
Year:
2022
Address:
Seattle, Washington
Venue:
SocialNLP
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
11–34
Language:
URL:
https://aclanthology.org/2022.socialnlp-1.2
DOI:
10.18653/v1/2022.socialnlp-1.2
Bibkey:
Cite (ACL):
Dana Ruiter, Thomas Kleinbauer, Cristina España-Bonet, Josef van Genabith, and Dietrich Klakow. 2022. Exploiting Social Media Content for Self-Supervised Style Transfer. In Proceedings of the Tenth International Workshop on Natural Language Processing for Social Media, pages 11–34, Seattle, Washington. Association for Computational Linguistics.
Cite (Informal):
Exploiting Social Media Content for Self-Supervised Style Transfer (Ruiter et al., SocialNLP 2022)
Copy Citation:
PDF:
https://aclanthology.org/2022.socialnlp-1.2.pdf
Code
 uds-lsv/3st
Data
CoLAGYAFC