Knowledgeable or Educated Guess? Revisiting Language Models as Knowledge Bases
arXiv:2106.09231
Abstract
Previous literatures show that pre-trained masked language models (MLMs) such as BERT can achieve competitive factual knowledge extraction performance on some datasets, indicating that MLMs can potentially be a reliable knowledge source. In this paper, we conduct a rigorous study to explore the underlying predicting mechanisms of MLMs over different extraction paradigms. By investigating the behaviors of MLMs, we find that previous decent performance mainly owes to the biased prompts which overfit dataset artifacts. Furthermore, incorporating illustrative cases and external contexts improve knowledge prediction mainly due to entity type guidance and golden answer leakage. Our findings shed light on the underlying predicting mechanisms of MLMs, and strongly question the previous conclusion that current MLMs can potentially serve as reliable factual knowledge bases.
arXiv commentAccepted to ACL2021(main conference)
References in corpus (7)
- Prefix-Tuning: Optimizing Continuous Prompts for Generation
- Assessing BERT's Syntactic Abilities
- Making Pre-trained Language Models Better Few-shot Learners
- Language Models are Open Knowledge Graphs
- Few-Shot Text Generation with Pattern-Exploiting Training
- Language Models as Few-Shot Learner for Task-Oriented Dialogue Systems
- Do Neural Language Representations Learn Physical Commonsense?