Pith. sign in

REVIEW 4 major objections 5 minor 64 references

Benchmarking Foundation Speech and Language Models for Alzheimer's Disease and Related Dementia Detection from Spontaneous Speech

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Whisper-medium embeddings outperform text and engineered features for three-way dementia classification from spontaneous speech.

desk verdict A useful foundation-model benchmark on a real-world ADRD dataset, but the comparison against traditional features is undermined by a binary-vs-three-class mismatch. read the letter →

arxiv 2506.11119 v1 pith:YJNQUHNN submitted 2025-06-09 cs.CL cs.SDeess.AS

Jingyu Li , Lingchao Mao , Hairong Wang , Zhendong Wang , Xi Mao , Xuelei Sherry Ni This is my paper
classification cs.CLcs.SDeess.AS
keywords Alzheimer'sdiseasedementiadetectionfoundationmodelsspeechembeddingsautomaticrecognitionWhisperspontaneousbenchmarking
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that frozen embeddings from open-source foundation speech models can serve as practical biomarkers for Alzheimer's disease and related dementias. On 1,189 spontaneous-speech recordings from the PREPARE dataset, it reports that Whisper-medium achieves the best three-way classification of healthy control, mild cognitive impairment, and Alzheimer's disease, at 0.731 accuracy and 0.802 AUC. It further claims that acoustic embeddings from automatic speech recognition models outperform text-embedding pipelines and handcrafted acoustic features, while adding pause annotations consistently improves text models. The result matters because it points to a scalable, non-invasive screening route that avoids manual transcription and feature engineering.

What carries the argument

The load-bearing object is the frozen, mean-pooled encoder embedding of a pre-trained foundation model, treated as a feature vector for a small trained classifier. For the audio pipeline, Whisper converts 16 kHz audio into log-Mel spectrograms and its transformer encoder produces clip-level embeddings (1024 dimensions for Whisper-medium); for the text pipeline, BERT-family models embed transcripts in which pauses are encoded as punctuation marks (comma, period, ellipsis for short, medium, long pauses). What carries the argument is the claim that large-scale pre-training on diverse speech lets Whisper encode both semantic content and paralinguistic timing cues, so a single frozen representation outperforms task-specific feature engineering.

What would settle it

Re-run the identical frozen-embedding classifiers on all 1,646 PREPARE training recordings without the spontaneous-speech filter, or on an independent spontaneous-speech corpus with age- and sex-matched groups; if Whisper-medium's margin over BERT-with-pause disappears or flips, the acoustic-superiority claim fails. Alternatively, report per-class recall on the 81 MCI samples: if MCI recall is near zero, the reported accuracy is carried almost entirely by HC/AD separation.

Watch

Extended reading notes

Core claim

The central claim is that among the benchmarked models, Whisper-medium's encoder embeddings—mean-pooled across the clip and concatenated with age and sex before a single-layer feedforward classifier—give the best performance for HC/MCI/AD classification, with accuracy $0.7307 \pm 0.0202$ and AUC $0.8024 \pm 0.0143$ over five random splits. Among language models, BERT with pause markers reaches the top accuracy of $0.6622 \pm 0.0131$ and AUC of $0.7444 \pm 0.0136$, and adding pause markers improves every text-based configuration. The paper also shows that Whisper-medium beats both Whisper-large and Whisper-tiny, that traditional feature sets such as eGeMAPS reach levels comparable to several smaller speech models but not to Whisper-medium, and that the advantage of speech models is attributed to their ability to retain non-semantic cues such as pauses and prosody without explicit annotation.

Load-bearing premise

The benchmark's validity rests on the assumption that the DBSCAN clustering of BERT embeddings plus human review correctly identifies the 1,189 spontaneous-speech samples; if that curation is biased, every model comparison in the paper describes a different, filtered task rather than spontaneous speech detection.

Editorial extensions

If this is right

  • Acoustic-only screening is feasible: the best speech model needs no transcript and no manual pause annotation to reach 0.731 accuracy and 0.802 AUC on a three-way clinical task.
  • Model scale is not monotonic: Whisper-medium outperforms Whisper-large, suggesting an optimal size near 769M parameters for this task and dataset.
  • Non-semantic information carries signal: adding pause markers to transcripts improves every BERT-family configuration tested.
  • A practical telehealth or mobile workflow is within reach: frozen embeddings plus a 128-unit classification layer is computationally light enough for deployment.
  • Traditional feature pipelines still compete: eGeMAPS ($0.643$ accuracy, $0.767$ AUC) matches several self-supervised speech models but falls short of Whisper-medium.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper reports accuracy and AUC but not per-class recall; with MCI at only 6.81% of the sample, per-class sensitivity would clarify whether the model actually distinguishes MCI or mostly separates HC from AD.
  • The 83-sample exclusion by DBSCAN and human review means the benchmark covers a curated spontaneous-speech subset; running the same pipeline on the full multilingual set would test whether the ranking holds outside English Cookie Theft descriptions.
  • The pause-annotation gain in text models and Whisper's acoustic advantage may be two views of the same timing signal; an experiment with deliberately pause-perturbed audio could separate semantic from timing contributions.
  • A longitudinal extension is the natural next test: if embedding trajectories shift with disease progression, the same frozen encoder could track within-person change rather than only cross-sectional status.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper benchmarks a range of open-source foundation speech and language models for three-class (HC/MCI/AD) classification of spontaneous speech from the PREPARE dataset. It reports that Whisper-medium achieves the best audio-based performance (accuracy 0.731, AUC 0.802), that BERT with pause annotation is the best text-based model (accuracy 0.662, AUC 0.744), and that acoustic/ASR-derived embeddings outperform text-based and traditional feature pipelines. The experiments use five stratified train-test splits with a frozen embedding extractor and a trained classification layer, which is a sound protocol. However, the comparison with traditional feature pipelines is undermined by a mismatch in classification tasks, and one claim about pause annotation is contradicted by the paper's own table.

Significance. If the results hold, the paper provides a useful benchmark of foundation models on a large, publicly available dataset for ADRD detection, with practical implications for scalable screening. The multi-seed evaluation, frozen-embedding design, and use of demographic covariates are strengths. The comparison against traditional features is important for establishing the value of foundation models, but the current evidence for that comparison is invalid as presented. The paper also offers a clear description of the data filtering pipeline, though the human-evaluation step needs more transparency.

major comments (4)
  1. [Section 3, Table 4; Supplementary S2; Section 4] The traditional feature results in Table 4 are produced by binary HC-vs-AD classification, as stated in Supplementary S2: "Each feature set was used independently to train binary classification models distinguishing Alzheimer's disease (AD) patients from cognitively normal controls." In contrast, Tables 2 and 3 report three-class HC/MCI/AD metrics. The accuracy and AUC values are therefore not comparable across these tables because the label sets, sample compositions, and class priors differ. The Discussion (Section 4) uses Table 4 to claim that "The Whisper-medium model significantly outperformed both the traditional feature-based models and other embedding models," which is a central conclusion of the paper. Please recompute the traditional baselines under the identical three-class protocol, or explicitly label Table 4 as a binary task and avoid direct numerical comparison. Without this correction, the claim that ASR embeddings outperform traditional feature pipelines is not supported.
  2. [Section 3, text around Table 2] The text states that "incorporating pause annotations consistently led to improved performance—in terms of both accuracy and AUC—compared to their counterparts without pause annotations." This is contradicted by the BioClinicalBERT row for Whisper-small transcriptions: with pause, AUC is 0.7118, while without pause it is 0.7126. This is the only exception in Table 2, but it invalidates the word "consistently" as used. Please qualify the claim, for example by noting that pause annotation improved both metrics in 11 of 12 configurations, or describe the exception explicitly.
  3. [Section 2.1, Figure S1] The identification of spontaneous speech samples relies on DBSCAN clustering of BERT embeddings followed by "human evaluation to classify clusters to language tasks," with 83 samples excluded as non-spontaneous. No details are given on the number of human evaluators, the annotation protocol, inter-rater agreement, or any validation of the cluster-to-task assignment. Because this filtering determines the composition of the final 1,189-sample benchmark, it is load-bearing for every model comparison. Please provide a more detailed description of the human evaluation step, including agreement measures, and consider a sensitivity analysis or a reproducible rule that does not depend on subjective judgment.
  4. [Section 2.3.2, Evaluation] The paper reports Area Under the Curve (AUC) for a three-class classification problem without specifying how multiclass AUC is computed. Please state whether macro-averaged one-vs-rest AUC (or another definition) is used. This is important for interpreting the absolute AUC values and for comparing with other studies.
minor comments (5)
  1. [Abstract] There is a typo: "A coustic-based approaches" should read "Acoustic-based approaches."
  2. [Figure 3 caption] The figure caption refers to panels "2(a)" and "2(b)" but the figure is numbered 3; this should be corrected to avoid confusion.
  3. [Section 3, Table 3 row names] The row label "Whisper-larger" in the text should be "Whisper-large" to match the model names and avoid ambiguity.
  4. [Table 2, column header] The column labeled "Data sizea" is described as "additional words trained" but the values for BERT (3.3B), BioBERT (18B), and BioClinicalBERT (0.5B) are not clearly sourced; please clarify whether these are pretraining corpus sizes and cite the corresponding references.
  5. [Data Availability] The paper states that data is publicly available but does not mention code. Given the complexity of the preprocessing and evaluation pipeline, releasing code would substantially improve reproducibility; please add a code availability statement or explain why it is omitted.

Circularity Check

0 steps flagged · score 0.0 of 10

Independent empirical benchmark against external labels; no circularity found.

full rationale

The paper makes no derived prediction that reduces to its inputs. It benchmarks frozen pretrained speech and language embeddings against held-out cognitive-status labels from the publicly available PREPARE dataset. The classification layer is trained on the training split and evaluated on a stratified held-out test split across five seeds, so the reported accuracies and AUCs are not fitted to the test labels. Pause-annotation thresholds and the use of Whisper for transcription are adopted from prior external work (Yuan et al., Radford et al.) and are not outputs of this study's fitted model. The DBSCAN/BERT-based filtering of spontaneous speech is a data-curation step applied before any cognitive-status prediction and does not use the labels, so it is not circular even though it is a subjective preprocessing choice. There are no self-citations that carry a load-bearing argument; citations to prior work are external. One apparent inconsistency—Supplementary S2 states the traditional feature pipeline trains binary AD-versus-HC classifiers, while Tables 2 and 3 report three-class metrics—is a benchmarking comparability concern, not circularity: the traditional-feature numbers are not used to construct the foundation-model results, and all pipelines are evaluated on real held-out labels. Therefore the central benchmark comparison is self-contained against external data and no circular step is present.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central benchmark rests on external labels from PREPARE/DementiaBank, the correctness of the spontaneous-speech filtering, and the choice of frozen-embedding plus single-layer classifier as the evaluation protocol. These are domain assumptions rather than fitted quantities; the only hand-chosen numeric parameters are the pause thresholds and classifier hyperparameters.

free parameters (3)
  • Word pause threshold = 0.05 seconds
    Pause markers inserted when inter-word gap exceeds 0.05 seconds; a hand-chosen threshold from prior work, affects text embeddings.
  • Pause bin boundaries = 0.5s and 2s
    Pauses categorized as short, medium, and long; values from Yuan et al. [29], influence the pause-aware BERT results.
  • Classification head hyperparameters = hidden 128, LR 5e-4, batch 32, patience 5
    Single-layer FFN settings chosen by authors; these affect all reported metrics but are not fitted to the test set.
assumptions (4)
  • domain assumption PREPARE phase 2 labels (HC, MCI, AD) are clinically accurate ground truth.
    Labels come from DementiaBank/PREPARE; no independent verification in this paper.
  • domain assumption The DBSCAN clustering plus human evaluation correctly identifies spontaneous speech tasks.
    Section 2.1 and Figure S1; if this is wrong, the filtered set is not spontaneous speech.
  • domain assumption Mean-pooled embeddings from frozen foundation models retain sufficient acoustic and linguistic information for classification.
    Used throughout Section 2.2; no analysis of which layers or pooling choices are optimal.
  • standard math The five repeated 80/20 splits give unbiased performance estimates.
    Standard practice, but splits overlap across repetitions, so the reported SDs are not independent repeated-measures uncertainties.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Benchmarking Foundation Speech and Language Models for Alzheimer's Disease and Related Dementia Detection from Spontaneous Speech." pith.science (2026). https://pith.science/paper/YJNQUHNN

@misc{pith2026250611119,
  author       = {Pith},
  title        = {Pith review of: Benchmarking Foundation Speech and Language Models for Alzheimer's Disease and Related Dementia Detection from Spontaneous Speech},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YJNQUHNN}},
  note         = {Machine review of arXiv:2506.11119}
}
read the original abstract

Background: Alzheimer's disease and related dementias (ADRD) are progressive neurodegenerative conditions where early detection is vital for timely intervention and care. Spontaneous speech contains rich acoustic and linguistic markers that may serve as non-invasive biomarkers for cognitive decline. Foundation models, pre-trained on large-scale audio or text data, produce high-dimensional embeddings encoding contextual and acoustic features. Methods: We used the PREPARE Challenge dataset, which includes audio recordings from over 1,600 participants with three cognitive statuses: healthy control (HC), mild cognitive impairment (MCI), and Alzheimer's Disease (AD). We excluded non-English, non-spontaneous, or poor-quality recordings. The final dataset included 703 (59.13%) HC, 81 (6.81%) MCI, and 405 (34.06%) AD cases. We benchmarked a range of open-source foundation speech and language models to classify cognitive status into the three categories. Results: The Whisper-medium model achieved the highest performance among speech models (accuracy = 0.731, AUC = 0.802). Among language models, BERT with pause annotation performed best (accuracy = 0.662, AUC = 0.744). ADRD detection using state-of-the-art automatic speech recognition (ASR) model-generated audio embeddings outperformed others. Including non-semantic features like pause patterns consistently improved text-based classification. Conclusion: This study introduces a benchmarking framework using foundation models and a clinically relevant dataset. Acoustic-based approaches -- particularly ASR-derived embeddings -- demonstrate strong potential for scalable, non-invasive, and cost-effective early detection of ADRD.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

64 extracted references · 52 canonical work pages

  1. [1]

    Jack Jr, D.A

    C.R. Jack Jr, D.A. Bennett, K. Blennow, M.C. Carrillo, et al. NIA‐AA research framework: toward a biological definition of Alzheimer's disease. Alzheimer's & dementia. 2018. 14(4): p. 535-562. https://doi.org/10.1016/j.jalz.2018.02.018

    doi:10.1016/j.jalz.2018.02.018 2018

  2. [2]

    Petersen, R.O

    R.C. Petersen, R.O. Roberts, D.S. Knopman, B.F. Boeve, et al. Mild cognitive impairment: ten years later. Archives of neurology. 2009. 66(12): p. 1447-1455. https://doi.org/10.1001/archneurol.2009.266

    work page doi:10.1001/archneurol.2009.266 2009

  3. [3]

    Van Dyck, C.J

    C.H. Van Dyck, C.J. Swanson, P. Aisen, R.J. Bateman, et al. Lecanemab in early Alzheimer’s disease. New England Journal of Medicine. 2023. 388(1): p. 9-21. https://doi.org/10.1056/NEJMoa2212948

    doi:10.1056/nejmoa2212948 2023

  4. [4]

    Sims, J.A

    J.R. Sims, J.A. Zimmer, C.D. Evans, M. Lu, et al. Donanemab in early symptomatic Alzheimer disease: the TRAILBLAZER-ALZ 2 randomized clinical trial. Jama. 2023. 330(6): p. 512-527. https://doi.org/10.1001/jama.2023.13239

    arXiv 2023

  5. [5]

    Mattke, Z

    S. Mattke, Z. Shi, M. Hanson, S. Mitchell, et al. Estimated investment need to increase England’s capacity to diagnose eligibility for an Alzheimer’s treatment to G7 average capacity levels. The Journal of Prevention of Alzheimer's Disease

    work page

  6. [6]

    S. Roth, N. Burnie, I. Suridjan, J.T. Yan, et al. Current diagnostic pathways for Alzheimer’s disease: a cross-sectional real-world study across six countries. Journal of Alzheimer's Disease Reports. 2023. 7(1): p. 659-674. https://doi.org/10.3233/ADR230007

    work page doi:10.3233/adr230007 2023

  7. [7]

    Y. Liu, H. Jun, A. Becker, C. Wallick, et al. Detection rates of mild cognitive impairment in primary care for the United States Medicare population. The journal of prevention of Alzheimer's disease. 2024. 11(1): p. 7-12. https://doi.org/10.14283/jpad.2023.131

    work page doi:10.14283/jpad.2023.131 2024

  8. [8]

    Mitchell and M

    A.J. Mitchell and M. Shiri‐Feshki. Rate of progression of mild cognitive impairment to dementia–meta‐analysis of 41 robust inception cohort studies. Acta psychiatrica scandinavica. 2009. 119(4): p. 252-265. https://doi.org/10.1111/j.1600-0447.2008.01326.x

    arXiv 2009

Show all 64 references
  1. [9]

    Mattke, H

    S. Mattke, H. Jun, E. Chen, Y. Liu, et al. Expected and diagnosed rates of mild cognitive impairment and dementia in the US Medicare population: observational analysis. Alzheimer's Research & Therapy. 2023. 15(1): p. 128. https://doi.org/10.1186/s13195-023-01272-z

    2023 doi

  2. [10]

    Gilmore-Bykovskyi, Y

    A.L. Gilmore-Bykovskyi, Y. Jin, C. Gleason, S. Flowers-Benton, et al. Recruitment and retention of underrepresented populations in Alzheimer's disease research: a systematic review. Alzheimer's & Dementia: Translational Research & Clinical Interventions. 2019. 5: p. 751-770. h...

    2019 doi

  3. [11]

    Glymour and J.J

    M.M. Glymour and J.J. Manly. Lifecourse social conditions and racial and ethnic patterns of cognitive aging. Neuropsychology review. 2008. 18: p. 223-254. https://doi.org/10.1007/s11065-008-9064-z

    2008 doi

  4. [12]

    Beltrami, G

    D. Beltrami, G. Gagliardi, R. Rossini Favretti, E. Ghidoni, et al. Speech analysis by natural language processing techniques: a possible tool for very early detection of cognitive decline? Frontiers in aging neuroscience. 2018. 10: p. 369. https://doi.org/10.3389/fnagi.2018.00369

    2018

  5. [13]

    Shakeri and M

    A. Shakeri and M. Farmanbar. Natural language processing in Alzheimer's disease research: Systematic review of methods, data, and efficacy. Alzheimer's & Dementia: Diagnosis, Assessment & Disease Monitoring. 2025. 17(1): p. e70082. https://doi.org/10.1002/dad2.70082

    2025 doi

  6. [14]

    Y. Zhu, B. Tran, X. Liang, J.A. Batsis, et al. Towards interpretability of speech pause in dementia detection using adversarial learning. in ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). 2022. IEEE. https://doi.org/10.1109/...

    2022

  7. [15]

    Gómez-Zaragozá, S

    L. Gómez-Zaragozá, S. Wills, C. Tejedor-Garcia, J. Marín-Morales, et al. Alzheimer disease classification through asr-based transcriptions: Exploring the impact of punctuation and pauses. arXiv preprint arXiv:2306.03443. 2023

    2023 arXiv

  8. [16]

    Radford, J.W

    A. Radford, J.W. Kim, T. Xu, G. Brockman, et al. Robust speech recognition via large-scale weak supervision. in International conference on machine learning

  9. [17]

    Goodglass and E

    H. Goodglass and E. Kaplan. Boston diagnostic aphasia examination booklet. 1983: Lea & Febiger

    1983

  10. [18]

    Ester, H.-P

    M. Ester, H.-P. Kriegel, J. Sander and X. Xu. A density-based algorithm for discovering clusters in large spatial databases with noise. in kdd. 1996. https://cdn.aaai.org/KDD/1996/KDD96-037.pdf?source=post_page

    1996

  11. [19]

    Fraser, J.A

    K.C. Fraser, J.A. Meltzer and F. Rudzicz. Linguistic features identify Alzheimer’s disease in narrative speech. Journal of Alzheimer’s disease. 2015. 49(2): p. 407-

    2015

  12. [20]

    Meilán, F

    J.J.G. Meilán, F. Martínez-Sánchez, J. Carro, D.E. López, et al. Speech in Alzheimer's disease: can temporal and acoustic parameters discriminate dementia? Dementia and geriatric cognitive disorders. 2014. 37(5-6): p. 327-334. https://doi.org/10.1159/000356726

    2014 doi

  13. [21]

    König, A

    A. König, A. Satt, A. Sorin, R. Hoory, et al. Automatic speech analysis for the assessment of patients with predementia and Alzheimer's disease. Alzheimer's & Dementia: Diagnosis, Assessment & Disease Monitoring. 2015. 1(1): p. 112-124. https://doi.org/10.1016/j.dadm.2014.11.012

    2015 doi

  14. [22]

    Robin, J.E

    J. Robin, J.E. Harrison, L.D. Kaufman, F. Rudzicz, et al. Evaluation of speech- based digital biomarkers: review and recommendations. Digital biomarkers

  15. [23]

    Eyigoz, S

    E. Eyigoz, S. Mathur, M. Santamaria, G. Cecchi, et al. Linguistic markers predict onset of Alzheimer's disease. EClinicalMedicine. 2020. 28. https://doi.org/10.1016/j.eclinm.2020.100583

    2020

  16. [24]

    Harrell, S.S

    K.M. Harrell, S.S. Wilkins, M.K. Connor and J. Chodosh. Telemedicine and the evaluation of cognitive impairment: the additive value of neuropsychological assessment. Journal of the American Medical Directors Association. 2014. 15(8): p. 600-606. https://doi.org/10.1016/j.jamda...

    2014 doi

  17. [25]

    Cullum, L

    C.M. Cullum, L. Hynan, M. Grosch, M. Parikh, et al. Teleneuropsychology: Evidence for video teleconference-based neuropsychological assessment. Journal of the International Neuropsychological Society. 2014. 20(10): p. 1028-

    2014

  18. [26]

    Meilán, F

    J.J. Meilán, F. Martínez-Sánchez, J. Carro, J.A. Sánchez, et al. Acoustic markers associated with impairment in language processing in Alzheimer's disease. The Spanish journal of psychology. 2012. 15(2): p. 487-494. https://doi.org/10.5209/rev_sjop.2012.v15.n2.38859

    2012 doi

  19. [27]

    Rapcan, S

    V. Rapcan, S. D'arcy, N. Penard, I.H. Robertson, et al. The use of telephone speech recordings for assessment and monitoring of cognitive function in elderly people. in INTERSPEECH. 2009. https://www.researchgate.net/profile/Richard- Reilly-2/publication/221479287

    2009

  20. [28]

    Haulcy and J

    R.m. Haulcy and J. Glass. Classifying Alzheimer's disease using audio and text- based representations of speech. Frontiers in Psychology. 2021. 11: p. 624137. https://doi.org/10.3389/fpsyg.2020.624137

    2021

  21. [29]

    J. Yuan, X. Cai, Y. Bian, Z. Ye, et al. Pauses for detection of Alzheimer’s disease. Frontiers in Computer Science. 2021. 2: p. 624488. https://www.frontiersin.org/articles/10.3389/fcomp.2020.624488/pdf

    2021

  22. [30]

    Y. Zhu, A. Obyat, X. Liang, J.A. Batsis, et al. Wavbert: Exploiting semantic and non-semantic speech using wav2vec and bert for dementia detection. in Interspeech. 2021. https://doi.org/10.21437/interspeech.2021-332

    2021 doi

  23. [31]

    2024 [accessed April 15 2024]

    https://www.drivendata.org/competitions/299/competition-nih-alzheimers- acoustic-2/. 2024 [accessed April 15 2024]

    2024

  24. [32]

    Lanzi, A.K

    A.M. Lanzi, A.K. Saylor, D. Fromm, H. Liu, et al. DementiaBank: Theoretical rationale, protocol, and illustrative analyses. American Journal of Speech- Language Pathology. 2023. 32(2): p. 426-438. https://doi.org/10.1044/2022_AJSLP-22-00281

    2023 doi

  25. [33]

    Becker, F

    J.T. Becker, F. Boiler, O.L. Lopez, J. Saxton, et al. The natural history of Alzheimer's disease: description of study cohort and accuracy of diagnosis. Archives of neurology. 1994. 51(6): p. 585-594. https://doi.org/10.1001/archneur.1994.00540180063015

    1994

  26. [34]

    DementiaBank. 2025. https://dementia.talkbank.org

    2025

  27. [35]

    Devlin, M.-W

    J. Devlin, M.-W. Chang, K. Lee and K. Toutanova. Bert: Pre-training of deep bidirectional transformers for language understanding. in Proceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technologies,...

    2019

  28. [36]

    Eyben, K.R

    F. Eyben, K.R. Scherer, B.W. Schuller, J. Sundberg, et al. The Geneva minimalistic acoustic parameter set (GeMAPS) for voice research and affective computing. IEEE transactions on affective computing. 2015. 7(2): p. 190-202. https://ieeexplore.ieee.org/document/7160715

    2015

  29. [37]

    Schuller, S

    B. Schuller, S. Steidl, A. Batliner, A. Vinciarelli, et al. The INTERSPEECH 2013 computational paralinguistics challenge: Social signals, conflict, emotion, autism

    2013

  30. [38]

    Abdul and A.K

    Z.K. Abdul and A.K. Al-Talabani. Mel frequency cepstral coefficient and its applications: A review. IEEE Access. 2022. 10: p. 122136-122158. https://ieeexplore.ieee.org/iel7/6287639/9668973/09955539.pdf

    2022

  31. [39]

    Pennebaker, M.E

    J.W. Pennebaker, M.E. Francis and R.J. Booth. Linguistic inquiry and word count: LIWC 2001. Mahway: Lawrence Erlbaum Associates. 2001. 71(2001): p. 2001. http://downloads.liwc.net.s3.amazonaws.com/LIWC2015_OperatorManual.pdf

    2001

  32. [40]

    Schroter, A.N

    H. Schroter, A.N. Escalante-B, T. Rosenkranz and A. Maier. DeepFilterNet: A low complexity speech enhancement framework for full-band audio based on deep filtering. in ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). 2022. IEE...

    2022 arXiv

  33. [41]

    J. Lee, W. Yoon, S. Kim, D. Kim, et al. BioBERT: a pre-trained biomedical language representation model for biomedical text mining. Bioinformatics. 2020. 36(4): p. 1234-1240. https://doi.org/10.1093/bioinformatics/btz682

    2020 doi

  34. [42]

    Alsentzer, J.R

    E. Alsentzer, J.R. Murphy, W. Boag, W.-H. Weng, et al. Publicly available clinical BERT embeddings. arXiv preprint arXiv:1904.03323. 2019. https://arxiv.org/pdf/1904.03323

    1904 arXiv

  35. [43]

    Baevski, Y

    A. Baevski, Y. Zhou, A. Mohamed and M. Auli. wav2vec 2.0: A framework for self-supervised learning of speech representations. Advances in neural information processing systems. 2020. 33: p. 12449-12460. https://proceedings.neurips.cc/paper/2020/file/92d1e1eb1cd6f9fba3227870bb6...

    2020

  36. [44]

    W.-N. Hsu, B. Bolte, Y.-H.H. Tsai, K. Lakhotia, et al. Hubert: Self-supervised speech representation learning by masked prediction of hidden units. IEEE/ACM transactions on audio, speech, and language processing. 2021. 29: p. 3451-

    2021

  37. [45]

    C. Wang, Y. Wu, Y. Qian, K. Kumatani, et al. Unispeech: Unified speech representation learning with labeled and unlabeled data. in International Conference on Machine Learning. 2021. PMLR. http://proceedings.mlr.press/v139/wang21y/wang21y.pdf

    2021

  38. [46]

    S. Chen, C. Wang, Z. Chen, Y. Wu, et al. Wavlm: Large-scale self-supervised pre-training for full stack speech processing. IEEE Journal of Selected Topics in Signal Processing. 2022. 16(6): p. 1505-1518. https://arxiv.org/pdf/2110.13900

    2022

  39. [47]

    Baevski, W.-N

    A. Baevski, W.-N. Hsu, Q. Xu, A. Babu, et al. Data2vec: A general framework for self-supervised learning in speech, vision and language. in International conference on machine learning. 2022. PMLR. https://proceedings.mlr.press/v162/baevski22a/baevski22a.pdf

    2022

  40. [48]

    Panayotov, G

    V. Panayotov, G. Chen, D. Povey and S. Khudanpur. Librispeech: an asr corpus based on public domain audio books. in 2015 IEEE international conference on acoustics, speech and signal processing (ICASSP). 2015. IEEE. http://www.danielpovey.com/files/2015_icassp_librispeech.pdf

    2015

  41. [49]

    Haider, S

    F. Haider, S. De La Fuente and S. Luz. An assessment of paralinguistic acoustic features for detection of Alzheimer's dementia in spontaneous speech. IEEE Journal of Selected Topics in Signal Processing. 2019. 14(2): p. 272-281. https://signalprocessingsociety.org/publications...

    2019

  42. [50]

    L. Tóth, G. Gosztolya, V. Vincze, I. Hoffmann, et al. Automatic Detection of mild cognitive impairment from spontaneous speech using ASR. in INTERSPEECH

  43. [51]

    Fraser, F

    K.C. Fraser, F. Rudzicz, N. Graham and E. Rochon. Automatic speech recognition in the diagnosis of primary progressive aphasia. in Proceedings of the fourth workshop on speech and language processing for assistive technologies. 2013. https://aclanthology.org/W13-3909/

    2013

  44. [52]

    S. Luz, F. Haider, S. de la Fuente Garcia, D. Fromm, et al. Alzheimer's dementia recognition through spontaneous speech. 2021. Frontiers Media SA. p. 780169. https://www.frontiersin.org/journals/computer- science/articles/10.3389/fcomp.2021.780169/pdf

    2021

  45. [53]

    McFee, C

    B. McFee, C. Raffel, D. Liang, D.P. Ellis, et al. librosa: Audio and music signal analysis in python. SciPy. 2015. 2015: p. 18-24. https://www.academia.edu/download/40296500/librosa.pdf

    2015

  46. [54]

    M. Ali. PyCaret: An open source, low-code machine learning library in Python

  47. [63]

    https://www.pycaret.org

    PyCaret version 1.0.0. https://www.pycaret.org. Tables and Figures Figure 1. MCI and ADRD Progression and Detection. We illustrate the life-course risk factors for dementia, the progression from healthy aging to mild cognitive impairment (MCI) and Alzheimer’s disease (AD), and...

  48. [64]

    For linguistic analysis, we used Linguistic Inquiry and Word Count (LIWC) to extract psychologically meaningful lexical features [39]

    were computed to characterize the spectral envelope of the speech signal. For linguistic analysis, we used Linguistic Inquiry and Word Count (LIWC) to extract psychologically meaningful lexical features [39]. All features were standardized prior to modeling. Classification Mod...

  49. [422]

    https://doi.org/10.3233/JAD-150520

    doi

  50. [1033]

    https://doi.org/10.1017/S1355617714000873

    doi

  51. [2013]

    https://eprints.gla.ac.uk/93665/1/93665.pdf

  52. [2015]

    https://www.isca-archive.org/interspeech_2015/toth15_interspeech.pdf

  53. [2020]

    4(3): p. 99-108. https://doi.org/10.1159/000510820

    doi

  54. [2023]

    https://proceedings.mlr.press/v202/radford23a.html

    PMLR. https://proceedings.mlr.press/v202/radford23a.html

  55. [2024]

    1022-1029

    11(4): p. 1022-1029. https://doi.org/10.14283/jpad.2024.24

    2024 doi

  56. [3460]

    https://arxiv.org/pdf/2106.07447

    arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.