What question did this study set out to answer?

The aim is to create a unified framework to facilitate adversarial attacks and improve NLP models.

January 1, 2020Open Access

TextAttack: A Framework for Adversarial Attacks, Data Augmentation, and Adversarial Training in NLP

View Full Paper

Ask AI

Bookmark

Authors

This framework enhances NLP model performance through adversarial attacks and data augmentation, democratizing access for researchers.

The aim is to create a unified framework to facilitate adversarial attacks and improve NLP models.
Developed a Python framework called TextAttack for adversarial attacks in NLP.
Incorporated 16 adversarial attack implementations with a modular design for flexibility.
Included modules for data augmentation and adversarial training.
TextAttack allows easy construction of attacks from various components, enhancing model robustness.
Supports a range of models, including BERT and transformers, across all GLUE tasks.
Facilitates rapid experimentation in adversarial training and data augmentation.

Morris et al. (2020) studied this question.

View Full Paper

Ask AI

Bookmark