Abstraction not Memory: BERT and the English Article System

TL;DR

This paper compares BERT and native speakers on English article prediction, revealing BERT's superior performance and suggesting it captures human-like generalizations rather than mere memorization.

Contribution

It demonstrates that BERT outperforms humans in article prediction and provides evidence that BERT's understanding aligns more with human intuition than simple memorization.

Findings

BERT outperforms humans in article prediction tasks.

BERT is especially better at detecting zero articles.

BERT's predictions align more with annotators than with the corpus when agreement is high.

Abstract

Article prediction is a task that has long defied accurate linguistic description. As such, this task is ideally suited to evaluate models on their ability to emulate native-speaker intuition. To this end, we compare the performance of native English speakers and pre-trained models on the task of article prediction set up as a three way choice (a/an, the, zero). Our experiments with BERT show that BERT outperforms humans on this task across all articles. In particular, BERT is far superior to humans at detecting the zero article, possibly because we insert them using rules that the deep neural model can easily pick up. More interestingly, we find that BERT tends to agree more with annotators than with the corpus when inter-annotator agreement is high but switches to agreeing more with the corpus as inter-annotator agreement drops. We contend that this alignment with annotators, despite being trained on the corpus, suggests that BERT is not memorising article use, but captures a high level generalisation of article use akin to human intuition.