Interpreting Pretrained Contextualized Representations via Reductions to Static Embeddings

ACL 2020