Refine speech recognition for cognitive assessments

There are multiple cognitive assessments that rely on spoken responses. This includes tasks where you name objects in a category, listen and repeat word lists or sentences, or tell stories. Currently, for all of these assessments scoring is performed by a trained professional. Automatic scoring can greatly improve and expand the use of this important class of assessments. The 3C platform uses browser-based speech recognition. This approach however requires refinement and optimization to ensure high accuracy. This project is to collect data from tasks relying on spoken responses, assess accuracy of the speech recognition through comparison with audio recordings, refine the protocols to maximize accuracy.

REB approval (Project Overview: H-10-23-9755) is already obtained.

This project involves:

• Understanding/learning about the aims/goals of various cognitive assessments
• Understanding/learing about psychometric properties of cognitive assessments
• Participant recruitment via social media, flyers, word of mouth
• Data screening, identifying complete data
• Identifying the limitations and areas of improvement for the use of speech recognition in cognitive assessments