Structural Persistence in Language Models: Priming as a Window into Abstract Language Representations
This work addresses the problem of understanding abstract language representations in AI models for researchers in computational linguistics and AI, offering an incremental tool for probing model capacities.
The study investigated whether modern neural language models exhibit structural priming, where sentence structures influence subsequent ones, to assess their ability to learn abstract structural information. It found that Transformer models show evidence of priming, with generalizations modulated by semantic information, and introduced a novel metric and Prime-LM corpus for controlled analysis.
We investigate the extent to which modern, neural language models are susceptible to structural priming, the phenomenon whereby the structure of a sentence makes the same structure more probable in a follow-up sentence. We explore how priming can be used to study the potential of these models to learn abstract structural information, which is a prerequisite for good performance on tasks that require natural language understanding skills. We introduce a novel metric and release Prime-LM, a large corpus where we control for various linguistic factors which interact with priming strength. We find that Transformer models indeed show evidence of structural priming, but also that the generalisations they learned are to some extent modulated by semantic information. Our experiments also show that the representations acquired by the models may not only encode abstract sequential structure but involve certain level of hierarchical syntactic information. More generally, our study shows that the priming paradigm is a useful, additional tool for gaining insights into the capacities of language models and opens the door to future priming-based investigations that probe the model's internal states.