Putting Words in BERT's Mouth: Navigating Contextualized Vector Spaces with Pseudowords
This work addresses the interpretability of BERT's internal representations for NLP researchers, though it is incremental as it builds on existing probing methods.
The researchers tackled the problem of understanding how contextualized vector spaces in models like BERT correspond to word senses by developing a method using pseudowords to explore these spaces. They found substantial regularity with distinct sense regions but also occasional 'sense voids' that lack intelligible meaning.
We present a method for exploring regions around individual points in a contextualized vector space (particularly, BERT space), as a way to investigate how these regions correspond to word senses. By inducing a contextualized "pseudoword" as a stand-in for a static embedding in the input layer, and then performing masked prediction of a word in the sentence, we are able to investigate the geometry of the BERT-space in a controlled manner around individual instances. Using our method on a set of carefully constructed sentences targeting ambiguous English words, we find substantial regularity in the contextualized space, with regions that correspond to distinct word senses; but between these regions there are occasionally "sense voids" -- regions that do not correspond to any intelligible sense.