Arborist accurately expands large-scale taxonomies with edge semantics that are heterogeneous and unobserved.
Arborist learns global latent representations of the edge semantics with local taxonomic-role embeddings of the nodes.
Arborist guarantees easy human verification by minimizing the network distance between predicted and actual taxonomy parents.
Emaad Manzoor, Dhananjay Shrouty, Rui Li, Jure Leskovec. Expanding Taxonomies with Implicit Edge Semantics. In Proceedings of The Web Conference (WWW). 2020.