LLM enriched speech features improve early Alzheimer detection accuracy
LLM-Anchored Paralinguistic Enrichment for Alzheimer's Disease Detection
Computation and LanguageSound
Summary
Detecting Alzheimer's disease early can be done by analyzing how people speak, since the disease changes both word choice and speech patterns like pauses. The researchers created a method called LAPE that combines language AI models with detailed speech cues, like pauses and stretched words, to get a better understanding. They tested their approach on standard Alzheimer speech datasets and found it outperforms previous tools. This way, speech analysis can become a more reliable, non-invasive way to screen for the disease early on.
What this means in practice
- •For clinical speech technology teams: Build more accurate Alzheimer's screening tools by integrating speech pattern features with language model analyses.
- •For digital health app developers: Enhance cognitive assessment apps with enriched speech and text features for early detection of cognitive decline.$Commercial implications: Enables commercial development of improved screening apps for healthcare providers and patients based on combined speech and language analysis.
Authors
Xiao Wei, Yuqin Lin, Yaru Cao, Jinyu Li, Bin Wen, Kai Li, Yueying Chen, Longbiao Wang, Jianwu Dang
Abstract
Speech-based automatic detection of Alzheimer's disease (AD) provides a non-invasive and scalable approach to early cognitive screening. AD affects both lexical-semantic organization and speech production, including atypical pauses and word elongations. However, existing methods have yet to fully integrate these paralinguistic cues with linguistic content. We propose LLM-Anchored Paralinguistic Enrichment (LAPE), which enriches LLM-derived linguistic representations with paralinguistic cues through three coordinated innovations. The first is prosodic event textualization, which enables the LLM to model pauses and elongations jointly with lexical content by encoding them as explicit markers with bounded duration-aware repetition. The second is lexico-prosodic unitization and chunking, which preserves event identity and magnitude in both modalities by pooling only consecutive word units. The third is text-anchored paralinguistic fusion, which integrates local and utterance-level speech features by using NormGate to normalize and dynamically scale them relative to text. We evaluate LAPE on ADReSS and ADReSSo using participant-level cross-validation and leave-one-subject-out evaluation. LAPE achieves state-of-the-art performance across all four primary settings. Code will be released upon acceptance.