Style Transfer from Non-Parallel Text by Cross-Alignment

Abstract

This paper focuses on style transfer on the basis of non-parallel text. This is an instance of a broad family of problems including machine translation, decipherment, and sentiment modification. The key challenge is to separate the content from other aspects such as style. We assume a shared latent content distribution across different text corpora, and propose a method that leverages refined alignment of latent representations to perform style transfer. The transferred sentences from one style should match example sentences from the other style as a population. We demonstrate the effectiveness of this cross-alignment method on three tasks: sentiment modification, decipherment of word substitution ciphers, and recovery of word order.

arXiv commentNIPS 2017 camera-ready. Added human evaluation on sentiment transfer

References in corpus (11)

• Categorical Reparameterization with Gumbel-Softmax
• Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks
• Convolutional Neural Networks for Sentence Classification
• The Concrete Distribution: A Continuous Relaxation of Discrete Random Variables
• Unsupervised Image-to-Image Translation Networks
• Unsupervised Cross-Domain Image Generation
• SeqGAN: Sequence Generative Adversarial Nets with Policy Gradient
• GANS for Sequences of Discrete Elements with the Gumbel-softmax Distribution
• Maximum-Likelihood Augmented Discrete Generative Adversarial Networks
• Toward Controlled Generation of Text
• Boundary-Seeking Generative Adversarial Networks

Cited by in corpus (108)

• The Curious Case of Neural Text Degeneration
• Augmented CycleGAN: Learning Many-to-Many Mappings from Unpaired Data
• Multimodal Unsupervised Image-to-Image Translation
• Unsupervised Machine Translation Using Monolingual Corpora Only
• DialogWAE: Multimodal Response Generation with Conditional Wasserstein Auto-Encoder
• Style Transfer as Unsupervised Machine Translation
• BaitWatcher: A lightweight web interface for the detection of incongruent news headlines
• Exploring the Role of Visual Content in Fake News Detection
• Discover Your Social Identity from What You Tweet: a Content Based Approach
• Multi-Task Neural Models for Translating Between Styles Within and Across Languages
• Controllable Unsupervised Text Attribute Transfer via Editing Entangled Latent Representation
• A Probabilistic Formulation of Unsupervised Text Style Transfer
• SGM: Sequence Generation Model for Multi-label Classification
• Multiple-Attribute Text Style Transfer
• Unsupervised Text Style Transfer using Language Models as Discriminators
• Transformer-based Conditional Variational Autoencoder for Controllable Story Generation
• Evaluating prose style transfer with the Bible
• Adversarially Regularized Autoencoders
• Style Transfer for Texts: Retrain, Report Errors, Compare with Rewrites
• ReviewRobot: Explainable Paper Review Generation based on Knowledge Synthesis
• CycleGT: Unsupervised Graph-to-Text and Text-to-Graph Generation via Cycle Training
• Spherical Latent Spaces for Stable Variational Autoencoders
• Educating Text Autoencoders: Latent Representation Guidance via Denoising
• Style Transfer in Text: Exploration and Evaluation
• GenNI: Human-AI Collaboration for Data-Backed Text Generation
• Encoder-Agnostic Adaptation for Conditional Language Generation
• MALA: Cross-Domain Dialogue Generation with Action Learning
• Generative Latent Flow
• CGMH: Constrained Sentence Generation by Metropolis-Hastings Sampling
• Back to the Future: Unsupervised Backprop-based Decoding for Counterfactual and Abductive Commonsense Reasoning
• CoCon: A Self-Supervised Approach for Controlled Text Generation
• Difficulty Controllable Generation of Reading Comprehension Questions
• Crossing You in Style: Cross-modal Style Transfer from Music to Visual Arts
• "Mask and Infill" : Applying Masked Language Model to Sentiment Transfer
• Fact-based Text Editing
• A Lexical, Syntactic, and Semantic Perspective for Understanding Style in Text
• Decomposing Textual Information For Style Transfer
• Deep Extrapolation for Attribute-Enhanced Generation
• Optimus: Organizing Sentences via Pre-trained Modeling of a Latent Space
• What is wrong with style transfer for texts?
• Style Transformer: Unpaired Text Style Transfer without Disentangled Latent Representation
• Sentiment Transfer using Seq2Seq Adversarial Autoencoders
• Improving GAN Training with Probability Ratio Clipping and Sample Reweighting
• Countering Online Hate Speech: An NLP Perspective
• PCAE: A Framework of Plug-in Conditional Auto-Encoder for Controllable Text Generation
• From Theories on Styles to their Transfer in Text: Bridging the Gap with a Hierarchical Survey
• Language Style Transfer from Sentences with Arbitrary Unknown Styles
• Domain Adaptive Text Style Transfer
• "Factual" or "Emotional": Stylized Image Captioning with Adaptive Learning and Attention
• $R^3$: Reverse, Retrieve, and Rank for Sarcasm Generation with Commonsense Knowledge
• Learning Neural Templates for Text Generation
• Grammatical Error Correction and Style Transfer via Zero-shot Monolingual Translation
• ER-AE: Differentially Private Text Generation for Authorship Anonymization
• Paraphrase Types for Generation and Detection
• Structuring Latent Spaces for Stylized Response Generation
• Data-to-Text Generation with Style Imitation
• Thank you BART! Rewarding Pre-Trained Models Improves Formality Style Transfer
• A Neural Approach to Irony Generation
• Implicit Deep Latent Variable Models for Text Generation
• Polite Dialogue Generation Without Parallel Data
• Counterfactual Story Reasoning and Generation
• Sequence Generation with Guider Network
• Plug and Play Autoencoders for Conditional Text Generation
• NLP Verification: Towards a General Methodology for Certifying Robustness
• APo-VAE: Text Generation in Hyperbolic Space
• Revision in Continuous Space: Unsupervised Text Style Transfer without Adversarial Learning
• Improving Disentangled Text Representation Learning with Information-Theoretic Guidance
• $A^{4}NT$: Author Attribute Anonymity by Adversarial Training of Neural Machine Translation
• Large-scale Hierarchical Alignment for Data-driven Text Rewriting
• A little goes a long way: Improving toxic language classification despite data scarcity
• Pre-train and Plug-in: Flexible Conditional Text Generation with Variational Auto-Encoders
• A Deep Factorization of Style and Structure in Fonts
• Variational Inference with Holder Bounds
• Review of Text Style Transfer Based on Deep Learning
• IMaT: Unsupervised Text Attribute Transfer via Iterative Matching and Translation
• Paraphrase Generation as Unsupervised Machine Translation
• Generating Diverse Story Continuations with Controllable Semantics
• Exploring Contextual Word-level Style Relevance for Unsupervised Style Transfer
• On Learning Text Style Transfer with Direct Rewards
• Investigation of Sentiment Controllable Chatbot
• Multi-type Disentanglement without Adversarial Training
• Unsupervised Stylish Image Description Generation via Domain Layer Norm
• Monolingual and Cross-lingual Zero-shot Style Transfer
• MSSRNet: Manipulating Sequential Style Representation for Unsupervised Text Style Transfer
• NeuraCrypt: Hiding Private Health Data via Random Neural Networks for Public Training
• Clickbait? Sensational Headline Generation with Auto-tuned Reinforcement Learning
• Text Modeling with Syntax-Aware Variational Autoencoders
• R&R: Metric-guided Adversarial Sentence Generation
• TextSETTR: Few-Shot Text Style Extraction and Tunable Targeted Restyling
• Unsupervised Neural Text Simplification
• Adversarial Decomposition of Text Representation
• Style Transfer Through Multilingual and Feedback-Based Back-Translation
• Neural Gaussian Copula for Variational Autoencoder
• Neural MultiVoice Models for Expressing Novel Personalities in Dialog
• HydraSum: Disentangling Stylistic Features in Text Summarization using Multi-Decoder Models
• Bag-of-Vectors Autoencoders for Unsupervised Conditional Text Generation
• QuaSE: Accurate Text Style Transfer under Quantifiable Guidance
• Molecular Attributes Transfer from Non-Parallel Data
• EAT: a simple and versatile semantic representation format for multi-purpose NLP
• (Male, Bachelor) and (Female, Ph.D) have different connotations: Parallelly Annotated Stylistic Language Dataset with Multiple Personas
• How Sequence-to-Sequence Models Perceive Language Styles?
• SE-DAE: Style-Enhanced Denoising Auto-Encoder for Unsupervised Text Style Transfer
• Learning to Represent Bilingual Dictionaries
• Finetuning Pretrained Transformers into Variational Autoencoders
• A Surprisingly Effective Fix for Deep Latent Variable Modeling of Text
• Compression, Transduction, and Creation: A Unified Framework for Evaluating Natural Language Generation
• XPROAX-Local explanations for text classification with progressive neighborhood approximation
• A Simple Geometric Method for Cross-Lingual Linguistic Transformations with Pre-trained Autoencoders