Through the open data analysis workpackage researchers at NCSR Demokritos led by Dr. Vassiliki Rentoumi created the CogAware Module. The module analyzes multi-lingual Alzheimer’s Disease (AD) patients’ language data to detect interesting language biomarkers for the early prognosis – diagnosis of AD. More specifically, researchers dealt with the evaluation of the transcripts of spoken language samples produced by the participants undertaking the standard Cookie Theft picture description task, originating from the Boston Diagnostic Aphasia Examination (Kaplan, 1983).

The goal is to automate the risk assessment of AD through the computational linguistic analysis of each sample text and without using any further knowledge regarding the patient (age, gender, clinical history, scores on other AD related tests etc.). The functionality and methodology comprising such a system are detailed in the following sections.

The data (speech transcripts) employed for training and testing the CogAware module were taken from three sources. Two from archived (English) language resources and one (Greek) collected within the context of iASiS. Participants were shown the “cookie theft” picture and were asked to describe what they could see happening. Regarding the English language resources two data sets were used, namely

CogAware module analyzes multi-lingual Alzheimer’s Disease (AD) patients’ language data to detect interesting language biomarkers for the early prognosis – diagnosis of AD. More specifically, CogAware performs the evaluation of the transcripts of spoken language samples produced by the participants undertaking the standard Cookie Theft picture description task.