Suicidal Risk Detection for Military Personnel

Park, Sungjoon; Park, Kiwoong; Ahn, Jaimeen; Oh, Alice

doi:10.18653/v1/2020.emnlp-main.198

Cited by 3 publications

(2 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Moreover, datasets from participants with verified diagnoses are typically very limited in size (as expected), such as the dataset released by Yaneva et al (2017) with 27 texts, or Rauschenberger et al (2016) with 47 texts. Albeit smaller, the resources using data from verified participants are comparatively more diverse in terms of the languages (English, Spanish, German) and conditions (autism, dyslexia, general mental health) they cover; by contrast, the social media resources that are publicly available typically target generic mental health related to anxiety, depression or suicide ideation and focus almost predominantly on English-language sources, notable exceptions being Park et al (2020) for Korean and Lee et al (2020) for Cantonese.…”

Section: Datasets and Resourcesmentioning

confidence: 99%

Natural Language Processing for Mental Disorders: An Overview

Calixto¹,

Yaneva²,

Cardoso³

2022

Natural Language Processing in Healthcare

View full text Add to dashboard Cite

Section: Datasets and Resourcesmentioning

confidence: 99%

Natural Language Processing for Mental Disorders: An Overview

Calixto¹,

Yaneva²,

Cardoso³

2022

Natural Language Processing in Healthcare

View full text Add to dashboard Cite

“…In realizing FedTherapist, the challenge remains to effectively capture mental health-related signals from a large corpus of spoken and typed user language on smartphones, which differs from prior NLP mental health studies based on social media (Yates et al, 2017;Park et al, 2020) -see Appendix G. To address such a challenge, we propose Context-Aware Language Learning (CALL) methodology, which integrates various temporal contexts of users (e.g., time, location) captured on smartphones to enhance the model's ability to sense mental health signals from the text data. Our evaluation of 46 participants shows that FedTherapist with CALL achieves more accurate mental health prediction than the model trained with non-text data (Wang et al, 2018), achieving 0.15 AUROC improvement and 8.21% reduction in MAE.…”

Section: Introductionmentioning

confidence: 99%

FedTherapist: Mental Health Monitoring with User-Generated Linguistic Expressions on Smartphones via Federated Learning

Shin,

Yoon,

Lee

et al. 2023

Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing

View full text Add to dashboard Cite

Psychiatrists diagnose mental disorders via the linguistic use of patients. Still, due to data privacy, existing passive mental health monitoring systems use alternative features such as activity, app usage, and location via mobile devices. We propose FedTherapist, a mobile mental health monitoring system that utilizes continuous speech and keyboard input in a privacy-preserving way via federated learning. We explore multiple model designs by comparing their performance and overhead for FedTherapist to overcome the complex nature of on-device language model training on smartphones. We further propose a Context-Aware Language Learning (CALL) methodology to effectively utilize smartphones' large and noisy text for mental health signal sensing. Our IRBapproved evaluation of the prediction of selfreported depression, stress, anxiety, and mood from 46 participants shows higher accuracy of FedTherapist compared with the performance with non-language features, achieving 0.15 AU-ROC improvement and 8.21% MAE reduction.

show abstract

Towards Building an Explainable System for Detecting Mental Illness with Evidence Support: Development and Usability Study (Preprint)

Kim¹,

Kang²,

Lee³

et al. 2023

Preprint

View full text Add to dashboard Cite

BACKGROUND Although mental illness is a leading cause of death, social stigma makes people hesitant to receive timely medical treatment. Online diagnosis systems can increase the accessibility of mental health care services. However, explaining illness detection results has been challenging in these systems because existing online diagnosis systems rely mainly on users’ self-assessments based on structured questionnaires. OBJECTIVE We proposed a system that provides detailed explanations (i.e., symptom evidence and solutions) about detected mental illnesses by analyzing users’ free-text inputs. METHODS Our system consists of three modules: a detection module, symptom extraction module, and solution suggestion module. The detection module analyzes the input text that the user describes, and accurately classifies the mental disorders that the user is likely to have. The system then extracts and delivers symptom evidence and solutions to explain the detection results better. A within-subject study was conducted to examine the effectiveness of the proposed system. Qualitative and quantitative data were collected using 36 questions, including seven open-ended questions, from 30 participants. RESULTS Analysis of the quantitative results demonstrated that additional symptom and solution information, along with the classification results, helped increase the explainability and usability of the online mental illness detection system. A total of 83.33% of the users preferred the system, which provided various explanations, and the free-text entry method helped them think more about themselves than answering structured questions. CONCLUSIONS The implementation of our proposed system revealed that (i) the system achieved fairly accurate performance in mental illness detection, (ii) providing explanations could improve user experience, and (iii) most users were satisfied with the system, while a few showed some concerns.

show abstract

Suicidal Risk Detection for Military Personnel

Cited by 3 publications

References 30 publications

Natural Language Processing for Mental Disorders: An Overview

Natural Language Processing for Mental Disorders: An Overview

FedTherapist: Mental Health Monitoring with User-Generated Linguistic Expressions on Smartphones via Federated Learning

Towards Building an Explainable System for Detecting Mental Illness with Evidence Support: Development and Usability Study (Preprint)

Contact Info

Product

Resources

About