Sat 22 Jun 2019 09:00 - 09:40 at 105A - Session 1

One of the key bottlenecks in building machine learning systems is creating and managing the massive training datasets that today’s models learn from. In this talk, I will describe our work on Snorkel (snorkel.stanford.edu), an open-source framework for building and managing training datasets, and describe three key operators for letting users build and manipulate training datasets: labeling functions, for labeling unlabeled data; transformation functions, for expressing data augmentation strategies; and slicing functions, for partitioning and structuring training datasets. These operators allow domain expert users to specify machine learning (ML) models via noisy operators over training data, leading to applications that can be built in hours or days, rather than months or years. I will describe recent work on modeling the noise and imprecision inherent in these operators, and using these approaches to train ML models that solve real-world problems, including a recent state-of-the-art result on the SuperGLUE natural language processing benchmark task.

Sat 22 Jun