Zipf’s law and the detection of the verbal evolution in Alzheimer's disease

Abstract

Background: In our society, it is undeniable the increase of prevalence of Alzheimer's Disease (AD) and of patients in different disease stages. Although it is essential to obtain an early diagnosis, it is also desirable to automatically detect the progression of the disease. Zipf’s law is a tool that allows, through the analysis of words frequency, to describe the linguistic evolution of patients with AD. Methods: A set of corpora of 20 patients with AD (10 GDS4 and 10 GDS5) and 10 controls, derived from three tests of oral production, have been used and studied statistically. Results: Deviations from the Zipf’s exponent in the words of mid-frequency for GDS5 patients have been observed, but not for GDS4. Discussion: Desviations on Zipf’s exponent in GDS5 versus control group show that it is possible to predict the evolution from one disease stage to another in the AD and determine when syntax is altered, exploring the simple oral production of the patient. In other words, variations in Zipf’s law can predict the syntactic evolution of these patients. Through future automatic detection systems we aim to describe the evolution of certain diseases with verbal alterations.