Low Resource Style Transfer via Domain Adaptive Meta Learning

TL;DR

This paper introduces DAML-ATM, a novel approach combining domain adaptive meta-learning and adversarial transfer for low-resource text style transfer, achieving state-of-the-art results on multi-domain datasets.

Contribution

It proposes a new meta-learning framework and an adversarial transfer model to improve style transfer in low-resource, unseen domains.

Findings

Achieves state-of-the-art results on multi-domain datasets.

Generalizes well to unseen low-resource domains.

Reduces performance degradation in new domain adaptation.

Abstract

Text style transfer (TST) without parallel data has achieved some practical success. However, most of the existing unsupervised text style transfer methods suffer from (i) requiring massive amounts of non-parallel data to guide transferring different text styles. (ii) colossal performance degradation when fine-tuning the model in new domains. In this work, we propose DAML-ATM (Domain Adaptive Meta-Learning with Adversarial Transfer Model), which consists of two parts: DAML and ATM. DAML is a domain adaptive meta-learning approach to learn general knowledge in multiple heterogeneous source domains, capable of adapting to new unseen domains with a small amount of data. Moreover, we propose a new unsupervised TST approach Adversarial Transfer Model (ATM), composed of a sequence-to-sequence pre-trained language model and uses adversarial style training for better content preservation and style transfer. Results on multi-domain datasets demonstrate that our approach generalizes well on unseen low-resource domains, achieving state-of-the-art results against ten strong baselines.