The availability of large annotated data can be a critical bottleneck in
training machine learning algorithms successfully, especially when applied to
diverse domains. weak supervision offers a promising alternative by
accelerating the creation of labeled training data using domain-spe