PhD 2023F: Benjamin Ascoli
Predicting Brain Age through Entity-Centric Analysis of Narratives in Picture Description
Benjamin Ascoli
Date: 2023-09-15 / 3:30 ~ 4:30 PM
Location: MSC E306
Abstract
This study seeks to create advancements in the understanding of brain aging through an innovative, entity-centric approach to analyzing narratives derived from picture description tasks. Utilizing the "Cookie Theft" task and the Aging Brain Cohort dataset, the project aims to dissect linguistic patterns surrounding identified entities and associated verbs in the task image, thereby exploring potential correlations with indicators of brain aging encompassing both structural and functional facets. Leveraging the capabilities of GPT-4, an in-depth discourse analysis is undertaken. This novel method involves entity linking and span analysis, complemented by the examination of action and state verbs. Initial analyses suggest promising potential in unveiling intricate linguistic markers possibly associated with brain aging, setting the stage for predictive modeling endeavors. As the project unfolds, it holds the promise of significantly contributing to the neurolinguistics field by offering a fresh perspective on the complex relationships between linguistic patterns and brain aging, possibly leading to the creation of refined diagnostic tools and interventions in the future.