StarCoder:  A general neural ensemble technique to support traditional scholarship, illustrated with a study of the post-Atlantic slave trade

1. Abstract

We introduce StarCoder, a graph-aware neural autoencoder modell that can be easily used by traditional scholars to leverage current state-of-the-art machine learning architectures. StarCoder has the specific goals of working on a broad range of data without modification, while allowing computer science researchers to easily extend and specialize its behavior. We describe StarCoder and its early successes in support of an ongoing study of the post-Atlantic slave trade.