English

LM-CORE: Language Models with Contextually Relevant External Knowledge

Computation and Language 2022-08-16 v1 Machine Learning

Abstract

Large transformer-based pre-trained language models have achieved impressive performance on a variety of knowledge-intensive tasks and can capture factual knowledge in their parameters. We argue that storing large amounts of knowledge in the model parameters is sub-optimal given the ever-growing amounts of knowledge and resource requirements. We posit that a more efficient alternative is to provide explicit access to contextually relevant structured knowledge to the model and train it to use that knowledge. We present LM-CORE -- a general framework to achieve this -- that allows \textit{decoupling} of the language model training from the external knowledge source and allows the latter to be updated without affecting the already trained model. Experimental results show that LM-CORE, having access to external knowledge, achieves significant and robust outperformance over state-of-the-art knowledge-enhanced language models on knowledge probing tasks; can effectively handle knowledge updates; and performs well on two downstream tasks. We also present a thorough error analysis highlighting the successes and failures of LM-CORE.

Keywords

Cite

@article{arxiv.2208.06458,
  title  = {LM-CORE: Language Models with Contextually Relevant External Knowledge},
  author = {Jivat Neet Kaur and Sumit Bhatia and Milan Aggarwal and Rachit Bansal and Balaji Krishnamurthy},
  journal= {arXiv preprint arXiv:2208.06458},
  year   = {2022}
}

Comments

Published at Findings of NAACL, 2022

R2 v1 2026-06-25T01:40:32.011Z