Data-centric AI is an approach to machine learning where data acquisition is the main barrier to success. Deploying an out of the box machine learning model has never been easier. Acquiring the data to evaluate and train the model is a completely different story though. Weak supervision uses weak signals to generate noisy labelled data from unlabelled data. Let me guide you through the process of using skweak a weak supervision library for Natural Language Processing to make your dataset creation processes scale. I will show you how to generate some noisy data using skweak and then we will train a NLP model.