Pith. sign in

REVIEW 4 major objections 7 minor 43 references

Cross-Demographic Portability of Deep NLP-Based Depression Models

T0 review · 4 major / 7 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a deep NLP depression classifier trained on younger adults keeps most of its accuracy—AUC 0.82 to 0.76—when applied to seniors, with 0.81 for seniors whose symptoms are consistent.

desk verdict Useful portability data, but the headline AUC gap is not statistically grounded and the LM retraining ambiguity needs resolving before believing the portability claim. read the letter →

arxiv 2412.19070 v1 pith:HMEPHPH7 submitted 2024-12-26 cs.CL

Tomek Rutowski , Elizabeth Shriberg , Amir Harati , Yang Lu , Ricardo Oliveira , Piotr Chlebek This is my paper · ORCID
classification cs.CL
keywords depressiondetectionnaturallanguageprocessingdemographicportabilitytransferlearningspeech-basedscreeningPHQ-8agemismatchdeep
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

The paper asks whether a deep NLP model that screens for depression from transcribed speech can be moved from the population it was trained on—adults aged 18–65, mostly in their 20s and 30s—to a much older retirement-community population without retraining. It reports that the model, trained on the General Population corpus, achieves an AUC (area under the ROC curve) of 0.82 on held-out younger speakers and AUC=0.76 on senior speakers, a modest drop despite almost no age overlap. For seniors whose self-reported depression status stayed consistent across weekly sessions, AUC rises to 0.81, nearly matching the matched-age result. The authors conclude that language markers of depression learned from younger adults are largely portable across age, gender, and ethnicity, with performance declining gradually as test age increases.

What carries the argument

The key machinery is an AWD-LSTM-style language model fine-tuned with ULMFiT-style transfer learning: the model is first pretrained on general text, retrained as a language model on the unlabeled depression corpora, then augmented with classification layers and trained on the GP depression labels. Classification uses a binary PHQ-8 cut-off (score at or above 10 means depressed), and evaluation is reported as ROC AUC. The critical analytical device is the matched-collection design: the GP and SP corpora were gathered with the same app, similar questions, and financial incentives, so the age mismatch is the main visible demographic difference; the SP corpus is then split into 'consistent' and 'inconsistent' patients by whether their depression class changed across sessions.

What would settle it

Measure the third-party speech-to-text word-error rate on the SP sessions by age and recompute AUC using only correctly transcribed responses; if the senior AUC rises to the in-domain level after controlling for transcription errors, the age-portability claim is partly an artifact of ASR quality rather than language content.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that a deep transfer-learned NLP depression classifier trained exclusively on a younger adult corpus transfers to an age-mismatched senior corpus with only a 0.06–0.07 absolute AUC loss (0.828 in-domain vs 0.761 cross-domain). The loss nearly disappears for the subgroup of seniors whose PHQ-8-derived class is consistent across longitudinal sessions (AUC=0.81, comparable to in-domain performance), while seniors whose class fluctuates are much harder to classify (AUC=0.61). The same model generalizes across gender splits and across self-reported ethnicity subgroups in the younger corpus, with most subgroup AUCs in the 0.79–0.89 range.

Load-bearing premise

The load-bearing assumption is that the GP and SP corpora are comparable in everything except age—same app, similar question prompts, similar incentives—so that the measured AUC gap reflects demographic portability rather than differences in collection site, device, speech-to-text quality, or how seniors answered the questions.

Editorial extensions

If this is right

  • A deployed screening model could be offered to older users without retraining, accepting a roughly six-point AUC reduction in exchange for immediate coverage.
  • For patients who report stable symptoms across weeks, the model's cross-age accuracy is near in-domain, supporting the use of longitudinal consistency as a design signal for speech-based screening.
  • Subgroup results suggest no gender- or ethnicity-specific retraining is needed for the younger population, though Hispanic users showed the lowest AUC among the tested groups.
  • Performance degrades as test age rises within the senior corpus, so age-aware thresholds or speech-to-text improvements may be needed for the oldest users.
  • Inconsistent patients should be treated as a separate evaluation population; their low AUC (0.61) warns against interpreting a single-session screener output as a stable diagnosis.

Reading between the lines

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

  • If the portability is genuine, the same transfer should hold across other protocol-matched demographic mismatches, such as education level, dialect region, or comorbid conditions; a multi-site replication would test this directly.
  • The consistent/inconsistent gap may partly reflect label noise: PHQ-8 self-reports that swing across weeks are less reliable ground truth, so the lower AUC on inconsistent patients could overstate model failure and understate true symptom tracking.
  • A natural extension is fine-tuning on a small labeled senior sample; the paper tests only zero-shot transfer, so even a few hundred senior sessions might close much of the remaining age gap.
  • The authors note shorter SP responses (450 vs 800 words) and fewer responses per session for consistent users; this confound suggests a testable hypothesis that response length, not age, drives part of the degradation.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 7 minor

Summary. The paper studies whether a deep NLP-based depression classifier trained on a younger general-population corpus (GP) transfers to a senior population (SP) without retraining. Using proprietary Ellipsis Health speech data with PHQ-8 labels, the authors report AUC=0.828 on age-matched held-out GP speakers and AUC=0.761 on the SP corpus. They further split the SP data by self-report label consistency over sessions, reporting AUC=0.82 for consistent and 0.61 for inconsistent patients, and they provide subgroup analyses by age, gender, and ethnicity. The central claim is that the model degrades only modestly across age groups, implying good demographic portability.

Significance. The question addressed is practically important: if a depression classifier trained on younger adults can be applied to older adults with minimal retraining, it could substantially lower the cost of deploying speech-based screening tools. A strength of the study is its use of two real-world, clinically labeled corpora with a deliberate age mismatch and a shared collection protocol, which is rare in this literature. The paper also reports several subgroup analyses that go beyond typical single-corpus evaluations. However, the central quantitative claim currently rests on point estimates without any uncertainty quantification, and the text suggests that SP transcripts may have been included in an unsupervised language-model retraining step, which would undermine the 'unseen corpus' claim. The post-hoc consistency split, which is presented as a key finding, is also not statistically supported. As it stands, the paper is a useful case study but does not yet establish the portability claim at the level of rigor expected for a journal publication.

major comments (4)
  1. [Section 3 (Language model retraining)] The sentence 'For the depression prediction task, we retrain the language model on our depression corpora without using labels' appears, given Section 2, to include both GP and SP in the unsupervised LM fine-tuning stage. If SP transcripts were used in this retraining, then the SP test set is not truly held out, and the reported AUC=0.761 is not a clean measure of demographic portability. Please state explicitly whether SP text was used in the LM retraining; if it was, retrain the LM on GP only and re-evaluate on SP.
  2. [Section 4.2, Table 3 and Figure 3] The AUC values 0.828 and 0.761 are point estimates with no confidence intervals or significance tests. Given Table 1 reports 161 SP subjects and 39 dep+ subjects, the standard error of AUC is plausibly on the order of 0.04-0.05, which makes the observed drop of 0.067 within sampling variation. To support the claim of 'only modest degradation,' the paper should report bootstrap confidence intervals clustered by subject or an equivalent statistical test.
  3. [Section 4.3 (Performance by patient class consistency)] The consistent-versus-inconsistent split is post hoc, with 119 versus 42 patients, no confidence intervals, and no correction for multiple testing. The reported AUC gap of 0.82 versus 0.61 may reflect label noise from PHQ-8 threshold crossing or the large response-length difference (consistent users average 3.8 responses per session versus 5.5 for inconsistent users, as stated in this section) rather than a patient-level consistency property. Please provide uncertainty estimates and control for response length or label noise before interpreting this result as a portability finding.
  4. [Section 2 (Corpora)] The paper states that the two corpora 'needed to match in terms of how they were collected. Otherwise, demographic differences would be confounded with speech elicitation methodology.' Yet the GP and SP corpora differ in several ways beyond age: different recruitment site, shorter SP sessions (~450 vs. ~800 words), different numbers of responses per session (Table 1), and potentially different ASR conditions, which the paper itself raises as a possible explanation for age-related degradation in Section 4.4. The conclusion that the AUC gap is attributable to age therefore needs to address these collection differences, or the result should be framed as portability across the joint set of demographic and collection differences.
minor comments (7)
  1. [Section 2 (after Table 1)] The text says 'the mean number of responses is slightly higher (6.1 vs 5.2),' but Table 1 and the preceding paragraph give a GP mean of 4.5 responses per session; the '5.2' appears to be an error.
  2. [Section 2 (Corpus statistics)] The text says the GP corpus 'contains over 15k sessions,' while Table 1 lists 12,872 total GP sessions; please clarify whether the 15k figure refers to a different filtering stage or is a typo.
  3. [Section 4.1, Table 3] The comparison with the system in [34] is explicitly indirect and on different datasets; the sentence 'demonstrate lower RMSE than the results for the system in [34]' should be phrased as non-comparable context rather than a claim of superiority.
  4. [Section 4.4, Table 4] The AUC values of 1.0 and 0.984 for SP age groups with 12 and 17 subjects are uninterpretable and should be flagged as such rather than presented in a table alongside groups with meaningful sample sizes.
  5. [Section 3 (Model description)] The model description omits several details needed for reproducibility, including layer sizes, vocabulary size after processing, learning-rate schedules, and the number of fine-tuning epochs; please provide these or state where the configuration is documented.
  6. [Abstract and Section 5] The abstract reports AUC=0.76, Table 3 implies 0.761, and the conclusion says '0.06 absolute AUC' while the difference is 0.067; please use a consistent rounding convention.
  7. [Section 4.3 and elsewhere] Phrases such as 'we discovered, interestingly' and 'it is worth noting that' are informal for a journal article; please adopt a more neutral scientific style.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the portability claim is an empirical transfer measurement with explicit train/test separation, not a derivation from its own inputs.

full rationale

The paper's central claim is an empirical measurement: a deep NLP depression classifier is trained on labeled GP sessions and then evaluated on a separate SP corpus. This is not a derivation, and no equation defines the SP AUC in terms of the training objective or in terms of the GP test result. The train/test separation is stated explicitly: 'Train and test partitions contain no overlapping speakers' for GP, and for SP, 'Given the size of this corpus, we use it only for testing in this study.' The GP AUC=0.828 is an independent same-distribution evaluation, and the SP AUC=0.761 is a genuinely out-of-corpus measurement under the stated protocol. The only self-citation, reference [24], is used to note that an earlier, smaller version of the GP corpus was used in prior work; it is provenance information and is not load-bearing for the portability conclusion. The consistent/inconsistent patient split in Section 4.3 is a post-hoc subgroup analysis rather than a fitted parameter renamed as a prediction; it does not construct the result from its inputs. The absence of confidence intervals around the SP AUC is a statistical robustness concern, not circularity. One ambiguity is that Section 3 says the language model is retrained on 'our depression corpora' before classification, which could be read as including SP; if SP text were used in that unsupervised retraining, the SP evaluation would not be fully out-of-corpus. However, the paper later states that 'the NLP system is always trained on the GP corpus training data,' and no equation or explicit statement confirms that SP transcripts were used in the LM retraining step. This is therefore a data-integrity caveat rather than a demonstrated circular reduction. Overall, the paper's claimed portability result is an empirical finding that stands or falls on the data, not on a circular argument.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The central portability claim rests on the PHQ-8 labeling scheme, the matched-collection assumption, and ASR quality, none of which are independently verified. The model itself is a standard architecture with disclosed training philosophy but undisclosed hyperparameters.

free parameters (1)
  • Unspecified deep learning hyperparameters
    The paper does not report learning rates, dropout rates, layer sizes, or early stopping criteria. These are fitted to the GP corpus but not disclosed, preventing independent assessment of variance or replication.
assumptions (4)
  • domain assumption PHQ-8 self-report scores are a valid ground truth for depression status.
    The paper maps PHQ-8 scores to binary classes (Section 2) and treats the questionnaire as the gold standard, without external clinical validation.
  • domain assumption The binary cutoff PHQ-8 >= 10 defines depression.
    This cutoff is taken from the literature (ref [25]) and is applied uniformly, which is standard but still a modeling assumption.
  • domain assumption The two corpora are matched in collection design except for age.
    Section 2 states the datasets 'needed to match in terms of how they were collected' to avoid confounding, but no objective similarity metrics are provided.
  • domain assumption Automatic speech recognition (ASR) transcriptions are accurate enough for the NLP features to be meaningful.
    The paper mentions third-party ASR (Section 3) but does not analyze ASR error rates or their impact on model performance.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Cross-Demographic Portability of Deep NLP-Based Depression Models." pith.science (2026). https://pith.science/paper/HMEPHPH7

@misc{pith2026241219070,
  author       = {Pith},
  title        = {Pith review of: Cross-Demographic Portability of Deep NLP-Based Depression Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HMEPHPH7}},
  note         = {Machine review of arXiv:2412.19070}
}
read the original abstract

Deep learning models are rapidly gaining interest for real-world applications in behavioral health. An important gap in current literature is how well such models generalize over different populations. We study Natural Language Processing (NLP) based models to explore portability over two different corpora highly mismatched in age. The first and larger corpus contains younger speakers. It is used to train an NLP model to predict depression. When testing on unseen speakers from the same age distribution, this model performs at AUC=0.82. We then test this model on the second corpus, which comprises seniors from a retirement community. Despite the large demographic differences in the two corpora, we saw only modest degradation in performance for the senior-corpus data, achieving AUC=0.76. Interestingly, in the senior population, we find AUC=0.81 for the subset of patients whose health state is consistent over time. Implications for demographic portability of speech-based applications are discussed.

Figures

Figures reproduced from arXiv: 2412.19070 by the authors.

Figure 2
Figure 2. PHQ distribution [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 4
Figure 4. Because of the distribution of data given in Figure [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 38 canonical work pages

  1. [1]

    Depression causes a heavy economic and societal burden

    INTRODUCTION The incidence of depression is increasing globally, wit h an estimated 300 million current cases worldwide [1][2][3]. Depression causes a heavy economic and societal burden

    work page

  2. [2]

    concerns

    CORPORA Our work uses two corpora of American English speech collected by Ellipsis Health. Despite the availabili ty of common corpora used for shared evaluations on depression prediction [22][23], it was necessary to use our own datasets for two important reasons. First, our sets n eeded to match in terms of how they were collected. Otherwise , demograph...

    work page

  3. [3]

    Various methods can be used for tokenization [32][33]

    DEEP LEARNING MODEL: NLP WITH TRANSFER LEARNING Our prediction model uses a deep learning language model with transfer learning [28]. Various methods can be used for tokenization [32][33]. We use the spaCy tokenization library; each word is represented by a unique ID. Our corpus word dictionary contains over 20 000 individual tokens. Given the overall sho...

    work page

  4. [4]

    Because depression is under-diagnosed, there is a pressing need for efficient, cost-effective screening and monitoring [6]

    and is presently exacerbated by COVID-19 [5]. Because depression is under-diagnosed, there is a pressing need for efficient, cost-effective screening and monitoring [6]. Digital health applications can play an important ro le in remote approaches to screening and monitoring. For fully automated systems, natural-language-based applications are promising. S...

    work page

  5. [5]

    RESULTS AND DISCUSSION 4.1. Comparison to past work Given our use of large but proprietary datasets for both training and evaluation, it is useful to provide evidence of how our approaches perform on shared corpora. Because we did not have access to benchmark corpora, we provide the following indirect comparisons. They are on differ ent datasets but, nota...

    work page 2019

  6. [6]

    ACKNOWLEDGMENTS We thank David Lin, Mike Aratow, Tahmida Nazreen, Chloe Owen, and Mainul Mondal for support and contributions

    work page

  7. [7]

    CONCLUSIONS AND FUTURE WORK We found that a state-of-the-art depression classifier based on deep NLP and transfer learning showed excellent portability over age, gender, and ethnicity. Using two corpora almost non-overlapping in age but similar in collection design, we found only a small degradation in binary classification performance (0.06 absolute AUC)...

    work page

  8. [8]

    Emotion recognition on speech signals using machine learning,

    M. Ghai, S. Lal, S. Duggal and S. Manik, “Emotion recognition on speech signals using machine learning,” in Proc. Int. Conf. Big Data Analytics and Comput. Intell., p p. 34–39, 2017

    work page 2017

Show all 43 references
  1. [9]

    Burden of Depressive Disorders by Country, Sex, Age, and Year: Findings from the Global Burden of Disease Study 2010,

    A. J. Ferrari, et al., “Burden of Depressive Disorders by Country, Sex, Age, and Year: Findings from the Global Burden of Disease Study 2010,” PLoS Med., vol. 10, Nov. 2013, doi: 10.1371/journal.pmed.1001547

    2010 doi

  2. [10]

    Geneva, Switzerland: World Health Organization, 2017

    World Health Organization, Depression and Other Common Mental Disorders: Global Health Estimates . Geneva, Switzerland: World Health Organization, 2017

    2017

  3. [11]

    https://www.nimh.nih.gov

    The National Institute of Mental Health. https://www.nimh.nih.gov

  4. [12]

    Lay be liefs about causes of depression,

    C. Lauber, L. Falcato, C. Nordt and W. Rössler, “Lay be liefs about causes of depression,” Acta Psychiatr. Scand. Suppl., vol. S-108, pp. 96–99, 2003

    2003

  5. [13]

    Changes in sleep pattern, sense of time, and digital media use during COVID-19 lockdown in Italy,

    N. Cellini, N. Canale, G. Mioni and S. Costa, “Changes in sleep pattern, sense of time, and digital media use during COVID-19 lockdown in Italy,” J. Sleep Res., vol. 29, 2020, doi:10.1111/jsr.13074

    2020 doi

  6. [14]

    The psychometric properties of depression screening tools in primary healthcare settings: A systema tic review,

    S. El-Den, T. F. Chen, Y. L. Gan, E. Wong and C. L. O’Reilly, “The psychometric properties of depression screening tools in primary healthcare settings: A systema tic review,” J. Affect. Disord., vol. 225, pp. 503–522, 2018

    2018

  7. [15]

    Speech recognition by machines and humans,

    R. P. Lippmann, “Speech recognition by machines and humans,” Speech Commun., vol. 22, pp. 1–16, 1997

    1997

  8. [16]

    Adieu features? End-to-end speech emotion recognition using a deep convolutional recurrent network,

    G. Trigeorgis et al., “Adieu features? End-to-end speech emotion recognition using a deep convolutional recurrent network,” in Proc. Int. Conf. Acoust. Speech and Signal Process., pp. 5200–5204, 2016

    2016

  9. [17]

    Automated computerized analysis of speech in psychiatric disorders,

    A. S Cohen and B. Elvevag, “Automated computerized analysis of speech in psychiatric disorders,” Curr. Opin. Psychiatry, vol. 27, pp. 203–209, 2014

    2014

  10. [18]

    Deep Learning for Depression Detection of Twitter User s,

    A. H. Orabi, P. Buddhitha, M.H. Orabi and D. Inkpen. “Deep Learning for Depression Detection of Twitter User s,” in Proc. 5th Workshop Comput. Linguistics and Clin. Psychol.: From Keyboard to Clinic, pp. 88–97, 2018

    2018

  11. [19]

    Depression and Self- Harm Risk Assessment in Online Forums,

    A. Yates, A. Cohan and N. Goharian, “Depression and Self- Harm Risk Assessment in Online Forums,” in Proc. 2017 Conf. Empirical Methods in Natural Lang. Process., pp. 2968–2978, 2017. [Online]. Available: https://aclweb.org/anthology/D17-1322

    2017

  12. [20]

    Voices of fear and anxiety and sadness and depression: The effects of speech rate and loudness on fear and anxiety and sadness and depression,

    A. W. Siegman and S. Boyle, “Voices of fear and anxiety and sadness and depression: The effects of speech rate and loudness on fear and anxiety and sadness and depression,” J. Abnorm. Psychol., vol. 102, pp. 430–437, 1993

    1993

  13. [21]

    Detecting depression with Audio/Text Sequence Modeling of Interviews,

    T. Al Hanai, M. Ghassemi and J. Glass, “Detecting depression with Audio/Text Sequence Modeling of Interviews,” in Proc. Interspeech 2018, pp. 1716–1720, 2018

    2018

  14. [22]

    Hybrid Depression Classification and Estimation from Audio Video and Text Information,

    L. Yang, H. Sahli, X. Xia, E. Pei, M. C. Oveneke and D. Jiang, “Hybrid Depression Classification and Estimation from Audio Video and Text Information,” in Proc. 7th Annu. Workshop Audio/Visual Emotion Challenge, pp. 45–51, 2017

    2017

  15. [23]

    Detecting depression from facial actions and vocal prosody,

    J. Cohn et al., “Detecting depression from facial actions and vocal prosody,” in Proc. Int. Conf. Affect. Comput. and Intell. Interact., pp. 1–7, 2009

    2009

  16. [24]

    Optimizing Speech-Input Length for Speaker-Independent Depression Classification,

    T. Rutowski, A. Harati, Y. Lu and E. Shriberg, “Optimizing Speech-Input Length for Speaker-Independent Depression Classification,” in Proc. Interspeech 2019, pp. 3023–3027, 2019

    2019

  17. [25]

    The Age of Artificial Emotional Intelligence,

    D. Schuller and B. W. Schuller, “The Age of Artificial Emotional Intelligence,” Computer, vol. 51, pp. 38–46, 2018

    2018

  18. [26]

    Natural Language Processing Methods for Acoustic and Landmark Event-based Features in Speech-based Depression Detection,

    Z. Huang, J. Epps and D. Joachim, “Natural Language Processing Methods for Acoustic and Landmark Event-based Features in Speech-based Depression Detection,” IEEE J. Sel. Topics Signal Process., vol. 14, pp. 435–448, 2020

    2020

  19. [27]

    Towards an Integrated Understanding of Speaking Rate in Conversation,

    J. Yuan, M. Liberman and C. Cieri, “Towards an Integrated Understanding of Speaking Rate in Conversation,” presented at Interspeech 2006, Pittsburgh, PA, USA, Sept. 17–21, 2006

    2006

  20. [28]

    State of Mind: Classification through Self- reported Affect and Word Use in Speech,

    E. M. Rathner, Y. Terhorst, N. Cummins, B. Schuller and H. Baumeister, “State of Mind: Classification through Self- reported Affect and Word Use in Speech,” in Proc. Interspeech 2018, pp. 267–271, 2018

    2018

  21. [29]

    The Distress Analysis Interview Corp us of human and computer interviews,

    J. Gratch et al., “The Distress Analysis Interview Corp us of human and computer interviews,” in Proc. 9th Int. Conf. Lang. Resour. and Eval., pp. 3123–3128, 2014

    2014

  22. [30]

    Beyond Age

    decreases in value as age increases. Similarly, the dark line (count for speakers below the age threshold) incre ases as it collects additional speakers. Figure 4 Age threshold Analysis for SP Model performance degrades as the age threshold increases (dark squares). This means...

  23. [31]

    AVEC 2019 Workshop and Challenge: State-of-Mind, Detecting Depression with AI, and Cross- Cultural Affect Recognition,

    F. Ringeval et al., “AVEC 2019 Workshop and Challenge: State-of-Mind, Detecting Depression with AI, and Cross- Cultural Affect Recognition,” in Proc. 9th Int. Audio/Visual Emotion Challenge and Workshop, pp. 3–12, 2019

    2019

  24. [32]

    AVEC 2017: Real-life Depression and Affect Recognition Workshop and Challenge,

    F. Ringeval et al., “AVEC 2017: Real-life Depression and Affect Recognition Workshop and Challenge,” in Proc. 7th Annu. Workshop Audio/Visual Emotion Challenge., pp. 3–9, 2017

    2017

  25. [33]

    Optimal cut-off score for diagnosing depression with the Patient Health Questionnaire (PHQ-9): a meta-analysis,

    L. Manea, S. Gilbody and D. Mcmillan, “Optimal cut-off score for diagnosing depression with the Patient Health Questionnaire (PHQ-9): a meta-analysis,” CMAJ, vol. 184, 2011, doi: 10.1503/cmaj.110829

    2011 doi

  26. [34]

    Distributed Representations of Words and Phrases and their Compositionality,

    T. Mikolov, I. Sutskever, K. Chen, G. Corrado and J. Dean, “Distributed Representations of Words and Phrases and their Compositionality,” in Adv. Neural Inform. Process. Syst., pp. 3111–3119, 2013

    2013

  27. [35]

    GloVe: Global Vectors for Word Representation,

    J. Pennington, R. Socher and C. Manning, “GloVe: Global Vectors for Word Representation,” in Proc. 2014 Conf. Empirical Methods Natural Lang. Process., pp. 1532–1543, 2014

    2014

  28. [36]

    How Transferable are Neural Networks in NLP Applications?,

    L. Mou, Z. Meng, R. Yan, G. Li and Y. Xu, “How Transferable are Neural Networks in NLP Applications?,” Proc. 2016 Conf. Empirical Methods in Natural Lang. Process, pp. 479–489, 2016

    2016

  29. [37]

    Regularizing and Optimizing LSTM Language Models,

    S. Merity, N. S. Keskar and R. Socher, “Regularizing and Optimizing LSTM Language Models,” CoRR, 2017, arXiv:1708.02182

    2017 arXiv

  30. [38]

    Universal Language Model Fine- tuning for Text Classification,

    J. Howard and S. Ruder, “Universal Language Model Fine- tuning for Text Classification,” arXiv preprint, 2018, arXiv:1801.06146

    2018 arXiv

  31. [39]

    Pointer Sentinel Mixture Models,

    S. Merity, C. Xiong, J. Bradbury and R. Socher, "Pointer Sentinel Mixture Models," CoRR, 2016, arXiv:1609.07843

    2016 arXiv

  32. [40]

    Neural Machine Translation of Rare Words with Subword Units,

    R. Sennrich, B. Haddow and A. Birch, “Neural Machine Translation of Rare Words with Subword Units,” arXiv preprint, 2015, arXiv:1508.07909

    2015 arXiv

  33. [41]

    Subword Regularization: Improving Neural Network Translation Models with Multiple Subword Candidates,

    T. Kudo, “Subword Regularization: Improving Neural Network Translation Models with Multiple Subword Candidates,” in Proc. 56th Annu. Meeting Assoc. Comput. Linguistics, pp. 66–75, 2018

    2018

  34. [42]

    Automatic Assessment of Depression Fro m Speech via a Hierarchical Attention Transfer Network and Attention Autoencoders,

    Z. Zhao et al., "Automatic Assessment of Depression Fro m Speech via a Hierarchical Attention Transfer Network and Attention Autoencoders," in IEEE J. Sel. Topics Signal Process., vol. 14, pp. 423–434, 2020

    2020

  35. [43]

    An Experimental Comparison of Performance Measures for Classification,

    C. Ferri, J. Hernández-Orallo and R. Modroiu, “An Experimental Comparison of Performance Measures for Classification,” Pattern Recognit. Lett., vol. 30, pp. 27–38, 2009

    2009

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.