| 1 |
What Does BERT Look At? An Analysis of BERT's Attention ...
|
|
|
|
Abstract:
Large pre-trained neural networks such as BERT have had great recent success in NLP, motivating a growing body of research investigating what aspects of language they are able to learn from unlabeled data. Most recent analysis has focused on model outputs (e.g., language model surprisal) or internal vector representations (e.g., probing classifiers). Complementary to these works, we propose methods for analyzing the attention mechanisms of pre-trained models and apply them to BERT. BERT's attention heads exhibit patterns such as attending to delimiter tokens, specific positional offsets, or broadly attending over the whole sentence, with heads in the same layer often exhibiting similar behaviors. We further show that certain attention heads correspond well to linguistic notions of syntax and coreference. For example, we find heads that attend to the direct objects of verbs, determiners of nouns, objects of prepositions, and coreferent mentions with remarkably high accuracy. Lastly, we propose an ... : BlackBoxNLP 2019 ...
|
|
Keyword:
Computation and Language cs.CL; FOS Computer and information sciences
|
|
URL: https://arxiv.org/abs/1906.04341
https://dx.doi.org/10.48550/arxiv.1906.04341
|
|
BASE
|
|
Hide details
|
|
|
| 2 |
Natural chunk-and-pass language processing: Just another joint source-channel coding model?
|
|
|
|
BASE
|
|
Show details
|
|
|
| 3 |
Improving Coreference Resolution by Learning Entity-Level Distributed Representations ...
|
|
|
|
BASE
|
|
Show details
|
|
|
| 4 |
Inducing Domain-Specific Sentiment Lexicons from Unlabeled Corpora ...
|
|
|
|
BASE
|
|
Show details
|
|
|
| 5 |
Large-scale Analysis of Counseling Conversations: An Application of Natural Language Processing to Mental Health ...
|
|
|
|
BASE
|
|
Show details
|
|
|
| 6 |
Inducing Domain-Specific Sentiment Lexicons from Unlabeled Corpora
|
|
|
|
BASE
|
|
Show details
|
|
|
| 7 |
Large-scale Analysis of Counseling Conversations: An Application of Natural Language Processing to Mental Health
|
|
|
|
BASE
|
|
Show details
|
|
|
| 8 |
Social biases determine spatiotemporal sparseness of ciliate mating heuristics
|
|
|
|
BASE
|
|
Show details
|
|
|
|
|
