AugLy: Data Augmentations for Robustness

By Zoe Papakipos, Joanna Bitton. 2022

55 citations recorded by the source corpus.

We introduce AugLy, a data augmentation library with a focus on adversarial robustness. AugLy provides a wide array of augmentations for multiple modalities (audio, image, text,&video). These augmentations were inspired by those that real users perform on social media platforms, some of which were not already supported by existing data augmentation libraries. AugLy can be used for any purpose where data augmentations are useful, but it is particularly well-suited for evaluating robustness and systematically generating adversarial attacks. In this paper we present how AugLy works, benchmark it compared against existing libraries, and use it to evaluate the robustness of various state-of-the-art models to showcase AugLy's utility. The AugLy repository can be found at https://github.com/facebookresearch/AugLy.

Subfields: Artificial Intelligence, Sociology and Political Science, Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Adina Williams
2. Douwe Kiela
3. Eric Michael Smith
4. Mohit Bansal
5. Melanie Kambadur
6. Jason Aaron Edward Weston
7. Eleonora Presani
8. Carlos Escolano

In Sociology and Political Science

1. Melanie Kambadur
2. Adina Williams
3. Eleonora Presani
4. Eric Michael Smith
5. Marta Ruiz Costa-jussà
6. Douwe Kiela
7. Tanmay Gupta
8. Aniruddha Kembhavi

In Safety, Robustness & Interpretability

1. Eric Michael Smith
2. Adina Williams
3. Melanie Kambadur
4. Eleonora Presani
5. Douwe Kiela
6. Melissa Hall
7. Jude Fernandes
8. Rebecca Qian

Citations in the corpus

• “I’m sorry to hear that”: Finding New Biases in Language Models with a Holistic Descriptor Dataset (2022, background): “existing sentences from text sources and swap demographic terms heuristically (Zhao et al., 2019; Ma et al., 2021; Wang et al., 2021; Papakipos and Bitton, 2022), utilize handcrafted grammars (Renduchintala et al.”
• Visual Programming: Compositional visual reasoning without training (2022, methodology): “Note that the prompt includes a mapping of emojis to their names in the AugLy [22] library that is used to implement Emoji module.”
• "I'm sorry to hear that": finding bias in language models with a holistic descriptor dataset (2022, background): “and some even select existing sentences from text sources and swap demographic terms heuristically (Zhao et al., 2019; Ma et al., 2021; Wang et al., 2021; Papakipos and Bitton, 2022).”
• Toxicity in Multilingual Machine Translation at Scale (2022, background): “, 2021), and swapping demographic terms in existing text, either heuristically (Ma et al., 2021; Wang et al., 2021; Zhao et al., 2019; Papakipos and Bitton, 2022) or using trained neural language models (Qian et al.”
• A Survey on Data Augmentation for Text Classification (2021, unknown): “[159] is not as big for textual data augmentation methods, but can be used for multiple modalities (audio, image, text, & video).”
• Perturbation Augmentation for Fairer NLP (2022, background): “The perturber generations show higher BLEU (Papineni et al., 2002) and ROUGE scores than do AugLy (Papakipos and Bitton, 2022) and TextFlint (Wang et al., 2021), as well as lower Levenshtein distance 7 to the human generated perturbations (see Table 2).”