"Computational analysis of aphasic narrative speech: Are we there yet?", Steven Bedrick

Lecture in the C-STAR series, delivered by Steven Bedrick, PhD (Oregon Health Sciences University) on August 23rd, 2024.
Dr. Bedrick's webpage: www.bedrick.org
Abstract:
Computational analysis of aphasic narrative speech: Are we there yet?
Traditionally, clinical assessment of aphasia relies on structured activities such as confrontation naming. While effective, these tests are somewhat limited in ecological validity, and only imperfectly model the kinds of speech production involved in activities of daily living and inter-personal communication. Narrative language samples give a potentially richer view of a patient’s true linguistic status, but are notoriously difficult to collect and analyze in reliable and valid ways. In this talk, I will describe how recent advances in computational techniques for language modeling are enabling new approaches to the analysis of narrative samples of aphasic speech, and discuss the potential clinical implications of such tools.