paper-with-me

Papers

Tailor: Generating and Perturbing Text with Semantic Controls

2021-08-17 · ACL ARR November 2021 11 · Anonymous

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.

📄 PDF Abstract BibTeX

Code (0)

등록된 구현이 없습니다.

Tasks

Data AugmentationNatural Language InferenceStyle Transfer

Similar Papers 제목 키워드 기반

Tailor: Generating and Perturbing Text with Semantic Controls

2021-07-15 · ACL 2022 5 · Alexis Ross, Tongshuang Wu, Hao Peng, Matthew E. Peters 외

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 Generation

D-SCoRE: Document-Centric Segmentation and CoT Reasoning with Structured Export for QA-CoT Data Generation

2025-08-02 · Weibo Zhou, Lingbo Li, Shangsong Liang arxiv

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 Engineering

Synthetic Text Generation using Hypergraph Representations

2023-09-06 · Natraj Raman, Sameena Shah

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 Generation

SETA: Semantic-Aware Token Augmentation for Domain Generalization

2024-03-18 · Jintao Guo, Lei Qi, Yinghuan Shi, Yang Gao

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 Generalization

Robust Contrastive Learning Using Negative Samples with Diminished Semantics

2021-10-27 · NeurIPS 2021 12 · Songwei Ge, Shlok Mishra, Haohan Wang, Chun-Liang Li 외

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