Reproduce “DECT: Harnessing LLM-assisted Fine-Grained Linguistic Knowledge and Label-Switched and Label-Preserved Data Generation for Diagnosis of Alzheimer's Disease” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Tingyu Mo, Jacqueline C. K. Lam, Victor O.K. Li, Lawrence Y. L. Cheung
A framework named DECT leverages Large Language Models to improve Alzheimer's Disease diagnosis from spontaneous speech, achieving 90.48% accuracy by distilling linguistic markers, identifying cognitive-linguistic atoms, and generating synthetic data. This approach addresses limitations of noisy, unstructured speech data and scarcity of labeled samples.
Key insights
- Large Language Models (LLMs) are effective at distilling fine-grained linguistic knowledge and identifying salient Cognitive-Linguistic (CL) atoms from noisy spontaneous speech, providing more discriminative features for AD diagnosis.
- The Label-Switched Label-Preserved (LSLP) data generation strategy, facilitated by LLMs, successfully mitigates data scarcity and significantly enhances model robustness and generalization by creating diverse, challenging synthetic samples.
- Fusing LLM-extracted linguistic markers and CL atoms into a structured Atom-Marker Representation (AMR) provides a more potent and discriminative input for AD detection models compared to raw transcript embeddings.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.