Inducing Relational Knowledge from BERT
This work addresses the need for understanding and extracting relational knowledge from advanced language models, which is incremental as it builds on existing BERT capabilities.
The paper tackled the problem of whether pre-trained language models like BERT capture relational knowledge beyond standard word embeddings, by proposing a method to distill such knowledge from BERT, resulting in a model that predicts relational instances with improved accuracy.
One of the most remarkable properties of word embeddings is the fact that they capture certain types of semantic and syntactic relationships. Recently, pre-trained language models such as BERT have achieved groundbreaking results across a wide range of Natural Language Processing tasks. However, it is unclear to what extent such models capture relational knowledge beyond what is already captured by standard word embeddings. To explore this question, we propose a methodology for distilling relational knowledge from a pre-trained language model. Starting from a few seed instances of a given relation, we first use a large text corpus to find sentences that are likely to express this relation. We then use a subset of these extracted sentences as templates. Finally, we fine-tune a language model to predict whether a given word pair is likely to be an instance of some relation, when given an instantiated template for that relation as input.