Utilizing Linguistic and Acoustic features from Arabic Transcripts for Early Detecting Alzheimer’s Disease Using Different Machine Learning Algorithms

Abstract

Alzheimer’s disease (AD) is a progressive neurological condition for which early detection is essential for effective intervention and symptom management. Language changes often serve as early indicators of cognitive decline in AD patients, facilitating early diagnosis. Factors affecting AD detection precision include feature extraction methods, the number of attributes used for selection, and the classifiers employed. While linguistic changes have been extensively studied in English, research on local and low-resource languages such as Arabic has been limited. In this study, we addressed this gap by developing a novel Arabic dataset comprising transcripts from patients with AD and healthy controls. Several machine learning algorithms have been used to explore the potential of linguistic features for early AD detection. Our approach leveraged speech patterns as acoustic features, including the silence rate associated with AD, to classify patients and controls. Notably, our findings suggest that linguistics, including nouns, verbs, pronouns, prepositions, and acoustic features extracted from Arabic speech data, can effectively distinguish between individuals with and without AD symptoms, the presentation of the results of various state-of-the-art textual classification mechanisms based on machine learning techniques is provided along with the baseline outcomes. with the highest accuracy of 0.958 achieved by Support vector machine algorithm.