This research leverages large language models to detect linguistic markers of Alzheimer's Disease from patient-interviewer dialogues, creating a non-invasive diagnostic tool.

• Employs fine-grained linguistic knowledge extraction to identify cognitive decline indicators
• Utilizes innovative label-switched and label-preserved data generation techniques
• Filters noisy, ambiguous data to focus on relevant cognitive markers
• Offers potential for earlier, more accessible AD screening

Medical Impact: This approach could significantly transform early AD diagnosis by providing a low-cost, scalable screening method that identifies subtle language impairments before other symptoms become apparent, enabling earlier intervention.

DECT: Harnessing LLM-assisted Fine-Grained Linguistic Knowledge and Label-Switched and Label-Preserved Data Generation for Diagnosis of Alzheimer's Disease