A menagerie of text ‘attack’ libraries
In image classifiers, a huge library of image augmentation tools crop, rotate, and otherwise edit input images to add variety to the training data and make classifiers more robust.
It’s reasonable to think that a similar technique could improve text data, but editing a sentence is more than moving pixels in x/y/color space. Changing one word or one letter can render a sentence incoherent or flip its meaning.
Note: there is some evidence that nonsensical data can help with some tasks — in Does Pretraining for Summarization Require Knowledge Transfer?, authors Krishna, Bigham, and Lipton show this for text summarization, and Uber has experimented with ‘Generative Teaching Networks’ with unreadable digits. But there’s debate about whether this is trains a robust model.
Text mutation libraries can have very different functions, such as adversarial ‘attacks’ against models, or those which add data before training. I’m going to talk about all of them here under that ‘mutation’ label.
This is one of the best-known libraries, with 1,700+ stars. It’s hosted by the QData team at University of Virginia. They have a ‘model zoo’ which shows which pretrained models and metrics you can expect after running the permutations / attacks.
GitHub - QData/TextAttack: TextAttack 🐙 is a Python framework for adversarial attacks, data…
Generating adversarial examples for NLP models [TextAttack Documentation on ReadTheDocs] About * Setup * Usage * Design…
This library was recently updated by the Natural Language Processing Lab at Tsinghua University. There is integration with HuggingFace models and datasets, which implies it could support almost any English or Chinese transformer model, and also older NLTK / other classifiers.
GitHub — thunlp/OpenAttack: An Open-Source Package for Textual Adversarial Attack.
Documentation * Features & Uses * Usage Examples * Attack Models * Toolkit Design OpenAttack is an open-source…
EDA: Easy Data Augmentation
This is our first option where we’re not reading into the model, but editing text to diversify our training data. The paper’s authors come from Carnegie Mellon, Google, MILA, and Dartmouth. A Chinese implementation has more stars than this original repo.
The techniques are simple: synonyms, random insertions, random swaps, and random deletion.
GitHub - jasonwei20/eda_nlp: Data augmentation for NLP, presented at EMNLP 2019
For a survey of data augmentation in NLP, see this repository/this paper. This is the code for the EMNLP-IJCNLP paper…
Hosted by University of Pretoria, this augmentation library can replace words with synonyms from WordNet or Word2Vec, or run a sentence through a translator and back to vary it. I recently saw Dr. Vukosi Marivate is looking for collaborators to update gensim and add features.
GitHub - dsfsi/textaugment: TextAugment: Text Augmentation Library
TextAugment is a Python 3 library for augmenting text for natural language processing applications. TextAugment stands…
Code for a recent paper from researchers at Auburn University and UC Berkeley. They look for nearest-neighbors inside of the embeddings to generate similar text which confounds the model.
GitHub - gongzhitaao/adversarial-text: Generate adversarial text via gradient methods
This is the code for our paper https://arxiv.org/abs/1801.07175 We are focusing on generate adversarial texts for text…
Code for a recent paper from Facebook Research. They also use gradients to find words to defeat the classifier.
I found the generated text to not be so realistic, or different enough to be worth changing the output class?
GitHub - facebookresearch/text-adversarial-attack: Repo for arXiv preprint "Gradient-based…
Install HuggingFace dependences conda install -c huggingface transformers pip install datasets (Optional) For attacks…
This is a lesser-known library (12 stars) by Dillon Niederhut, but its
add_love permutation made the rounds on Twitter, and was funny enough that I made a note of the library. The idea is that sentiment analysis classifiers can easily confuse a negative sentence with a positive one, simply by ending it with ‘love’.
add_leet which swaps out characters.
GitHub - deniederhut/niacin: Enrich your data
A Python library for replacing the missing variation in your text data. Data collected for model training necessarily…
This article was written in October 2021. If my recommendations change in the future, I’ll update it on this GitHub README.