💡📉Dataset programming is no longer a problem
Snorkel - a framework for data programming. The approach of this framework is to use various heuristics and a priori knowledge to automatically label datasets. The project started at Stanford as a tool to help mark up datasets for the information extraction task, and now the developers are creating a platform for use by external customers.
Snorkel's arsenal includes three key tools:
-marking functions for creating a dataset;
-transforming functions for dataset augmentation;
-slicing functions that highlight subsets in the dataset that are critical for the performance of learning models.