Visualizing and Measuring the Geometry of BERT
Abstract
Transformer architectures show significant promise for natural language processing. Given that a single pretrained model can be fine-tuned to perform well on many different tasks, these networks appear to extract generally useful linguistic features. A natural question is how such networks represent this information internally. This paper describes qualitative and quantitative investigations of one particularly effective model, BERT. At a high level, linguistic features seem to be represented in separate semantic and syntactic subspaces. We find evidence of a fine-grained geometric representation of word senses. We also present empirical descriptions of syntactic representations in both attention matrices and individual word embeddings, as well as a mathematical argument to explain the geometry of these representations.
Cite
@article{arxiv.1906.02715,
title = {Visualizing and Measuring the Geometry of BERT},
author = {Andy Coenen and Emily Reif and Ann Yuan and Been Kim and Adam Pearce and Fernanda Viégas and Martin Wattenberg},
journal= {arXiv preprint arXiv:1906.02715},
year = {2019}
}
Comments
8 pages, 5 figures