'Layer su Layer': Identifying and Disambiguating the Italian NPN Construction in BERT's family
For linguists and NLP researchers, this work provides empirical evidence on how constructional information is encoded in PLMs for a lesser-studied language, though the findings are incremental.
This study evaluates how well BERT's contextual embeddings encode the Italian NPN construction, finding that constructional form and meaning are reflected across layers, contributing empirical evidence to the dialogue between constructionist theory and neural language models.
Interpretability research has highlighted the importance of evaluating Pretrained Language Models (PLMs) and in particular contextual embeddings against explicit linguistic theories to determine what linguistic information they encode. This study focuses on the Italian NPN (noun-preposition-noun) constructional family, challenging some of the theoretical and methodological assumptions underlying previous experimental designs and extending this type of research to a lesser-investigated language. Contextual vector representations are extracted from BERT and used as input to layer-wise probing classifiers, systematically evaluating information encoded across the model's internal layers. The results shed light on the extent to which constructional form and meaning are reflected in contextual embeddings, contributing empirical evidence to the dialogue between constructionist theory and neural language modelling