Tailor: Generating and Perturbing Text with Semantic Controls
Making controlled perturbations is essential for various tasks (e.g., data augmentation), but building task-specific generators can be expensive. We introduce Tailor, a task-agnostic generation system that perturbs text in a semantically-controlled way. With unlikelihood training, Tailor's generator is designed to follow a series of control codes derived from semantic roles. Through modifications of these control codes, Tailor can produce fine-grained perturbations. We implement a set of operations on control codes that can be composed into complex perturbation strategies, and demonstrate their effectiveness in three applications. First, Tailor facilitates the construction of high-quality contrast sets that are lexically diverse and less biased than original task test data. Second, paired with automated labeling heuristics, Tailor helps improve model generalization through data augmentation: we obtain an average gain of 1.73 on an (natural language inference) NLI challenge set by perturbing just $\sim5\%$ of training data. Third, without any finetuning overhead, Tailor's perturbations effectively improve compositionality in fine-grained style transfer, outperforming fine-tuned baselines on 5 transfers.
Code (0)
등록된 구현이 없습니다.
Tasks
Data AugmentationNatural Language InferenceStyle TransferSimilar Papers 제목 키워드 기반
Tailor: Generating and Perturbing Text with Semantic Controls
Controlled text perturbation is useful for evaluating and improving model generalizability. However, current techniques rely on training a model for every target perturbation, which is expensive and hard to generalize. W…
Data AugmentationDiversityStyle TransferText GenerationD-SCoRE: Document-Centric Segmentation and CoT Reasoning with Structured Export for QA-CoT Data Generation
The scarcity and high cost of high-quality domain-specific question-answering (QA) datasets limit supervised fine-tuning of large language models (LLMs). We introduce $\textbf{D-SCoRE}$, a training-free framework that le…
Prompt EngineeringSynthetic Text Generation using Hypergraph Representations
Generating synthetic variants of a document is often posed as text-to-text transformation. We propose an alternate LLM based method that first decomposes a document into semantic frames and then generates text using this…
Hypergraph representationsText GenerationSETA: Semantic-Aware Token Augmentation for Domain Generalization
Domain generalization (DG) aims to enhance the model robustness against domain shifts without accessing target domains. A prevalent category of methods for DG is data augmentation, which focuses on generating virtual sam…
Data AugmentationDomain GeneralizationRobust Contrastive Learning Using Negative Samples with Diminished Semantics
Unsupervised learning has recently made exceptional progress because of the development of more effective contrastive learning methods. However, CNNs are prone to depend on low-level features that humans deem non-semanti…
Contrastive LearningData AugmentationSelf-Supervised Learning