How Contextual are Contextualized Word Representations? Comparing the Geometry of BERT, ELMo, and GPT-2 Embeddings Kawin Ethayarajh author 2019-11 text Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP) Kentaro Inui editor Jing Jiang editor Vincent Ng editor Xiaojun Wan editor Association for Computational Linguistics Hong Kong, China conference publication ethayarajh-2019-contextual 10.18653/v1/D19-1006 https://aclanthology.org/D19-1006/ 2019-11 55 65