What’s so special about BERT’s layers? A closer look at the NLP pipeline in monolingual and multilingual models

Authors:
  • WVWietse de Vries
    Wietse de Vries
  • ACAndreas van Cranenburgh
    Andreas van Cranenburgh
  • MNMalvina Nissim
    Malvina Nissim

Abstract
Wietse de Vries, Andreas van Cranenburgh, Malvina Nissim. Findings of the Association for Computational Linguistics: EMNLP 2020. 2020.