Text Style Transfer: An Introductory Overview

TL;DR

Text Style Transfer (TST) addresses changing a text's stylistic attributes while preserving content. The paper surveys the problem, datasets, evaluation metrics, subtasks, and approaches, emphasizing challenges such as non-parallel data and style-content entanglement. It discusses supervised, non-parallel, and large-language-model–driven methods, along with automatic and human evaluation strategies and ethical considerations. By consolidating datasets, methods, and evaluation practices, the work provides a practical, foundational guide for researchers and practitioners entering the TST domain.

Abstract

Text Style Transfer (TST) is a pivotal task in natural language generation to manipulate text style attributes while preserving style-independent content. The attributes targeted in TST can vary widely, including politeness, authorship, mitigation of offensive language, modification of feelings, and adjustment of text formality. TST has become a widely researched topic with substantial advancements in recent years. This paper provides an introductory overview of TST, addressing its challenges, existing approaches, datasets, evaluation measures, subtasks, and applications. This fundamental overview improves understanding of the background and fundamentals of text style transfer.