Character-Level Linguistic Biomarkers for Precision Assessment of Cognitive Decline: A Symbolic Recurrence Approach

Abstract Early and accurate detection of Alzheimer’s disease (AD) remains a critical challenge for precision health. Traditional cognitive assessments often miss subtle, individualized patterns of decline, while conventional linguistic analyses focus on word-level features that may overlook fine-grained speech disruptions. We test the hypothesis that character-level features in speech transcripts capturing pauses, repetitions, and hesitations at the finest linguistic granularity can serve as novel biomarkers for cognitive decline, revealing personalized linguistic signatures that manifest uniquely in each individual. Our biomarker discovery framework employs symbolic character-level encoding followed by recurrence quantification analysis to transform speech transcripts into visual recurrence plots that reveal temporal speech dynamics. Siamese networks learn embeddings from these plots to capture discriminative patterns at the character level. We validate our hypothesis using the DementiaBank corpus, demonstrating that character-level biomarkers achieve superior discriminative capability compared to conventional word-level approaches (95.9% vs. 87.5% AUC), while providing interpretable recurrence plot visualizations. Our findings establish that character-level linguistic features contain significant biomarker information for cognitive assessment, representing a fundamental shift from word-based to character-based analysis for precision health applications in dementia screening. Competing Interest Statement The authors have declared no competing interest. Funding Statement This study was funded by the NSF Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: The study used ONLY openly available human data that were originally located at: https://dementia.talkbank.org/access/English/Pitt.html I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes Footnotes (e-mail: faisal.aqlan{at}louisville.edu). (e-mail: huiyang{at}psu.edu). Data Availability The data that support the findings of this study are available from the Pitt Corpus within DementiaBank, maintained by TalkBank at Carnegie Mellon University and the University of Pittsburgh School of Medicine. Access to the data in DementiaBank is password protected and restricted to members of the DementiaBank consortium group. In accordance with TalkBank rules, any use of data from this corpus must be accompanied by appropriate corpus references and acknowledgment of grant support (NIA AG03705 and AG05133). Interested researchers can request access to the Pitt Corpus by visiting https://dementia.talkbank.org/access/English/Pitt.html. Established researchers and clinicians working with dementia can contact TalkBank to request access credentials.