Language abnormalities in Alzheimer's disease indicate reduced informativeness.

Area of Science:

• Computational linguistics
• Neuropsychology
• Machine learning

Background:

• Alzheimer's Disease (AD) is associated with language abnormalities.
• Understanding the cognitive basis of these abnormalities is crucial for diagnosis.

Purpose of the Study:

• To investigate the cognitive underpinnings of language deficits in Alzheimer's Disease (AD) using a cross-linguistic computational approach.
• To enhance the understanding and diagnostic accuracy of AD through linguistic analysis.

Main Methods:

• Computational analysis of language samples from English and Persian speakers (AD patients and healthy controls).
• Extraction of language indicators specific to AD.
• Introduction of a machine learning metric, Language Informativeness Index (LII), to quantify empty speech.

Main Results:

• Consistent AD language indicators were found across English and Persian, despite structural differences.
• AD indicators from English achieved 90% accuracy in classifying AD in Persian, indicating cross-linguistic transferability.
• Significant correlations were observed between AD language indicators and empty speech, supporting a universal deficit in message generation.

Conclusions:

• Language impairments in AD originate from deficits in universal message formation, not language-specific structures.
• This research advances the understanding of psycholinguistic deficits in AD.
• The developed approach supports the creation of multilingual diagnostic tools, promoting health equity.