TextAttack: A Framework for Data Augmentation and Adversarial Training in NLP скачать в хорошем качестве

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

00:10:00 - Beginning of the talk ◾ Title: TextAttack: A Python Framework for Adversarial Attacks, Data Augmentation, and Adversarial Training in NLP Slides: https://www.dropbox.com/s/s7rr3fnfmnf... ◾ Speaker: Jack Morris ◾ Twitter: @jxmorris12 ◾ Talk Description TextAttack is a Python framework for adversarial attacks and data augmentation in NLP. TextAttack's adversarial attacks are constructed from four components: a goal function, transformation, constraints, and search method. Data augmenters are constructed from a transformation and constraints. TextAttack implements many options for each component that cover 16 papers from the literature. TextAttack is super easy to use – just install it, and everything you need will download automatically – and ships with 82 pre-trained models for a variety of tasks. ◾ About Speaker: Jack Morris is a researcher at the University of Virginia interested in computer programs that understand language and the world around us. He's a co-creator of TextAttack and has participated in other research projects related to the robustness of NLP models. He will begin the Google AI Residency in October. Outside of research, Jack likes cycling, reading books, and watching baseball. ◾ About dair.ai Website: https://dair.ai/ GitHub: https://github.com/dair-ai Twitter:   / dair_ai   Newsletter: https://dair.ai/newsletter/ Slack: https://join.slack.com/t/dairai/share... ◾ Code of Conduct: https://github.com/dair-ai/dair-ai.gi...