Rachel Ostrand, Research Staff Member, Cognitive Science IBM Research

Automated Assessment of Spontaneous Speech Production
 as a Predictive Marker of Cognitive Decline

ABSTRACT: Dementia and cognitive decline affect millions of people. Diagnosis is generally performed using clinical instruments, which can include brain imaging, invasive biological testing, and/or time-consuming behavioral batteries. In this talk, I’ll discuss the growing methodology of using linguistic characteristics calculated from participants’ spontaneous speech as a predictive marker of cognitive decline. I’ll present two experiments in which participants produced spontaneous speech samples, from which a suite of linguistic characteristics were automatically calculated and then used to predict their clinical cognitive scores. In addition, I’ll discuss the feasibility of scaling this methodology to a larger and non-laboratory based population.