A Closer Look at Linguistic Knowledge in Masked Language Models: The Case of Relative Clauses in American English
This work addresses the problem of understanding linguistic knowledge in language models for researchers, but it is incremental as it builds on existing probing methods with a specific focus.
The study evaluated BERT, RoBERTa, and ALBERT on their ability to handle relative clauses in American English, finding that while they perform well on grammatical probing, they show significant weaknesses in semantic knowledge that affect overall performance.
Transformer-based language models achieve high performance on various tasks, but we still lack understanding of the kind of linguistic knowledge they learn and rely on. We evaluate three models (BERT, RoBERTa, and ALBERT), testing their grammatical and semantic knowledge by sentence-level probing, diagnostic cases, and masked prediction tasks. We focus on relative clauses (in American English) as a complex phenomenon needing contextual information and antecedent identification to be resolved. Based on a naturalistic dataset, probing shows that all three models indeed capture linguistic knowledge about grammaticality, achieving high performance. Evaluation on diagnostic cases and masked prediction tasks considering fine-grained linguistic knowledge, however, shows pronounced model-specific weaknesses especially on semantic knowledge, strongly impacting models' performance. Our results highlight the importance of (a)model comparison in evaluation task and (b) building up claims of model performance and the linguistic knowledge they capture beyond purely probing-based evaluations.