Probing for targeted syntactic knowledge through grammatical error detection
This work addresses the need for better understanding of syntactic knowledge in language models, though it is incremental as it builds on existing probing methods.
The study tackled the problem of evaluating whether pre-trained language models robustly encode subject-verb agreement by using grammatical error detection as a diagnostic probe, finding that masked language models linearly encode relevant information while autoregressive models perform similarly to a baseline, but performance varies with training data and syntactic constructions.
Targeted studies testing knowledge of subject-verb agreement (SVA) indicate that pre-trained language models encode syntactic information. We assert that if models robustly encode subject-verb agreement, they should be able to identify when agreement is correct and when it is incorrect. To that end, we propose grammatical error detection as a diagnostic probe to evaluate token-level contextual representations for their knowledge of SVA. We evaluate contextual representations at each layer from five pre-trained English language models: BERT, XLNet, GPT-2, RoBERTa, and ELECTRA. We leverage public annotated training data from both English second language learners and Wikipedia edits, and report results on manually crafted stimuli for subject-verb agreement. We find that masked language models linearly encode information relevant to the detection of SVA errors, while the autoregressive models perform on par with our baseline. However, we also observe a divergence in performance when probes are trained on different training sets, and when they are evaluated on different syntactic constructions, suggesting the information pertaining to SVA error detection is not robustly encoded.