Do Transformers Parse while Predicting the Masked Word?

TL;DR

This paper demonstrates that moderate-sized transformers can approximate the Inside-Outside algorithm for parsing and that masked language models inherently capture parsing structures, explaining their ability to encode linguistic information.

Contribution

It shows that transformers can explicitly perform approximate parsing and that masked language models are implicitly biased towards algorithms like Inside-Outside for parsing tasks.

Findings

Transformers can approximate the Inside-Outside algorithm.

Masked language models capture parsing structures.

Explicit transformer constructions enable >70% F1 constituency parsing.

Abstract

Pre-trained language models have been shown to encode linguistic structures, e.g. dependency and constituency parse trees, in their embeddings while being trained on unsupervised loss functions like masked language modeling. Some doubts have been raised whether the models actually are doing parsing or only some computation weakly correlated with it. We study questions: (a) Is it possible to explicitly describe transformers with realistic embedding dimension, number of heads, etc. that are capable of doing parsing -- or even approximate parsing? (b) Why do pre-trained models capture parsing structure? This paper takes a step toward answering these questions in the context of generative modeling with PCFGs. We show that masked language models like BERT or RoBERTa of moderate sizes can approximately execute the Inside-Outside algorithm for the English PCFG [Marcus et al, 1993]. We also show that the Inside-Outside algorithm is optimal for masked language modeling loss on the PCFG-generated data. We also give a construction of transformers with $50$ layers, $15$ attention heads, and $1275$ dimensional embeddings in average such that using its embeddings it is possible to do constituency parsing with $>70\%$ F1 score on PTB dataset. We conduct probing experiments on models pre-trained on PCFG-generated data to show that this not only allows recovery of approximate parse tree, but also recovers marginal span probabilities computed by the Inside-Outside algorithm, which suggests an implicit bias of masked language modeling towards this algorithm.