Ideas

Can we correlate any longitudinal data with NACC?

Data

• dementia/English/Lanzi: Alyssa Lanzi’s new data

• dementia/English/Delaware

• What are the standard for acoustic features?

• Motor cortex/frontal control may also be impacted

• Vocal tremer

• What are the predictors? How automatic can we make it?

The DementiaBank Acoustics Project is a working title for an acoustic-only challenge for AD detection. This document serves as the lab notebook for this project.

This project will attempt to replicate some of the results of Wang 2019 and Martinc 2021, but focusing on minimizing human involvement; we will first work on raw transcript classification with ERNIE (cutting all CHAT annotations), then introduce pause-encoding in a manner similar to Yuan 2021 which is automated by MFA. Goal is to replicate the results of Yuan 2021/or even Martinc 2021 in a completely automated manner.

Suppose you have two non mutually exclusive sets \(E\) or \(F\).

DeMorgan’s Law:

\begin{equation} (E\ and\ F)^{C} = (E^{C}\ or\ F^{C}) \end{equation}

\begin{equation} (E\ or\ F)^{C} = (E^{C}\ and\ F^{C}) \end{equation}