Hierarchical representation in neural language models: Suppression and recovery of expectations

Publication
Proceedings of the 2019 ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP