MIT Press

Table 1

Overview of the survey.

Motivation	Data	Method	Extended Applications
Artistic writing Communication Mitigating social issues	Tasks Formality Toxicity Politeness Authorship Gender Simplicity Humor Sentiment Romance Topic Biasedness Political slant Key Properties Parallel vs. non-parallel Uni- vs. bi-directional Dataset size Large vs. small word overlap	On Parallel Data Multi-tasking Inference techniques Data augmentation On Non-Parallel Data Disentanglement Prototype editing Pseudo data construction	Helping Other NLP Tasks Paraphrasing Data augmentation Adversarial robustness Persona-consistent dialog Anonymization Summarization Style-specific MT

Motivation

Data

Method

Extended Applications

Artistic writing
Communication
Mitigating social issues

Tasks

Formality
Toxicity
Politeness
Authorship
Gender
Simplicity
Humor
Sentiment
Romance
Topic
Biasedness
Political slant

Key Properties

Parallel vs. non-parallel
Uni- vs. bi-directional
Dataset size
Large vs. small word overlap

On Parallel Data

Multi-tasking
Inference techniques
Data augmentation

On Non-Parallel Data

Disentanglement
Prototype editing
Pseudo data construction

Helping Other NLP Tasks

Paraphrasing
Data augmentation
Adversarial robustness
Persona-consistent dialog
Anonymization
Summarization
Style-specific MT