Content preserving text generation with attribute controls

Abstract

In this work, we address the problem of modifying textual attributes of sentences. Given an input sentence and a set of attribute labels, we attempt to generate sentences that are compatible with the conditioning information. To ensure that the model generates content compatible sentences, we introduce a reconstruction loss which interpolates between auto-encoding and back-translation loss components. We propose an adversarial loss to enforce generated samples to be attribute compatible and realistic. Through quantitative, qualitative and human evaluations we demonstrate that our model is capable of generating fluent sentences that better reflect the conditioning information compared to prior methods. We further demonstrate that the model is capable of simultaneously controlling multiple attributes.

NIPS 2018

Cited by in corpus (30)

• Controllable Unsupervised Text Attribute Transfer via Editing Entangled Latent Representation
• Educating Text Autoencoders: Latent Representation Guidance via Denoising
• Neural Language Generation: Formulation, Methods, and Evaluation
• Domain Adaptive Text Style Transfer
• From Theories on Styles to their Transfer in Text: Bridging the Gap with a Hierarchical Survey
• "My Way of Telling a Story": Persona based Grounded Story Generation
• The Daunting Task of Real-World Textual Style Transfer Auto-Evaluation
• Data-to-Text Generation with Style Imitation
• Effective writing style imitation via combinatorial paraphrasing
• Rethinking Text Attribute Transfer: A Lexical Analysis
• A Neural Approach to Irony Generation
• Neural Response Generation with Meta-Words
• Contextual Text Style Transfer
• Revision in Continuous Space: Unsupervised Text Style Transfer without Adversarial Learning
• Plug and Play Autoencoders for Conditional Text Generation
• Generating Syntactically Controlled Paraphrases without Using Annotated Parallel Pairs
• IMaT: Unsupervised Text Attribute Transfer via Iterative Matching and Translation
• Fine-grained Sentiment Controlled Text Generation
• Exploring Contextual Word-level Style Relevance for Unsupervised Style Transfer
• Multi-type Disentanglement without Adversarial Training
• Semi-supervised Text Style Transfer: Cross Projection in Latent Space
• SE-DAE: Style-Enhanced Denoising Auto-Encoder for Unsupervised Text Style Transfer
• Syntax Matters! Syntax-Controlled in Text Style Transfer
• Sentence Bottleneck Autoencoders from Transformer Language Models
• Empirical Evaluation of Supervision Signals for Style Transfer Models
• Reflection-based Word Attribute Transfer
• EAT: a simple and versatile semantic representation format for multi-purpose NLP
• Diamonds in the Rough: Generating Fluent Sentences from Early-Stage Drafts for Academic Writing Assistance
• Transforming Multi-Conditioned Generation from Meaning Representation
• A Simple Geometric Method for Cross-Lingual Linguistic Transformations with Pre-trained Autoencoders