LoRA-Tuned Large Language Models for Dementia Detection via Multi-View Speech-Derived Features
For clinical dementia screening, this work provides a unified multi-view LLM approach that outperforms prior single-view methods, though it is incremental as it applies existing LoRA tuning to a new task.
The paper introduces a LoRA-tuned LLM that integrates four speech-derived views (ASR transcripts with pauses, discourse cues, fluency stats, phonological sequences) for dementia detection, achieving 90.14% F1 on ADReSSo, with ablation confirming each view's contribution.
Early detection of dementia enables timely intervention, and reflecting cognitive impairment, spontaneous speech offers a non-invasive screening modality. Conventional approaches often focus on a single representational dimension -- such as acoustic descriptors, pause modeling, automatic speech recognition (ASR) transcripts, or multimodal fusion -- limiting integrative reasoning across heterogeneous cognitive symptoms. We propose a low-rank adaptation (LoRA)-tuned large language model (LLM) that performs structured multi-view reasoning over four complementary speech-derived signals: ASR transcripts with pause markers, discourse-level topic cues, temporal fluency statistics, and phonological sequences. These cues are encoded within a unified prompt, enabling a single LLM to learn a coherent decision function without modality-specific encoders or late-stage fusion. On ADReSSo, our best model achieves an F1-score of 90.14%, and ablation confirms the complementary contribution of each view.