Demonstration of panda

Entity matching (EM) refers to the problem of identifying tuple pairs in one or more relations that refer same real world entities. Supervised machine learning (ML) approaches, and deep based approaches particular, typically achieve state-of-the-art results. However, these require many labeled examples, form non-matching pairs, which are expensive time-consuming label. In this paper, we introduce Panda, a weakly supervised system specifically designed for EM. Panda uses labeling function abstraction as Snorkel, where functions (LF) user-provided programs can generate large amounts (somewhat noisy) labels quickly cheaply, then be combined via model accurate final predictions. To support users developing LFs EM, provides an integrated development environment (IDE) lives modern browser architecture. Panda's IDE facilitates development, debugging, life-cycle management context EM tasks, similar how IDEs such Visual Studio Eclipse excel general-purpose programming. includes novel features purpose-built smart data sampling, builtin library utility functions, automatically generated LFs, visual debugging finally, EM-specific model. We show demo greatly accelerate high-quality solutions using weak supervision.