We’re delighted to to have received a NSF award to develop computational models of preditive sentence comprehension in bilingualism. This project is a collaboration with Edith Kaan (U Florida) and Maria Teresa Bajo (U Grenada) (separate award).
The basic idea is this: We start with participants from three bilingual populations who listen to an audiobook story in both their first and second languages (much prior research is text-based, uses isolated sentences, and focuses on just one group of bilinguals.) Linguistic properties of the story are annotated using NLP tools to capture features like syntactic dependencies. These annotations quantify what participants might predict, word-by-word, and also how information is used and weighted to form those predictions. The quality of fit between these metrics and EEG data are compared across L1 and L2 neural datasets. We also examine statistical fit across models that differ in the amount of training data and language balance to evaluate how neural responses reflect multilingual language exposure.
There is a lot of work ahead of us and we’re excited to get started. We will be advertising soon for a post-doctoral fellow to join the project. Stay tuned for more!