File Download

There are no files associated with this item.

  Links for fulltext
     (May Require Subscription)

Article: Detecting suicide risk using knowledge-aware natural language processing and counseling service data

TitleDetecting suicide risk using knowledge-aware natural language processing and counseling service data
Authors
KeywordsOnline counseling services
Suicide prevention
Natural language processing
Knowledge graph
Artificial intelligence
Issue Date2021
PublisherPergamon. The Journal's web site is located at http://www.elsevier.com/locate/socscimed
Citation
Social Science & Medicine, 2021, v. 283, p. article no. 114176 How to Cite?
AbstractRationale: Detecting users at risk of suicide in text-based counseling services is essential to ensure that at-risk individuals are flagged and prioritized. Objective: The objective of this study is to develop a domain knowledge-aware risk assessment (KARA) model to improve our ability of suicide detection in online counseling systems. Methods: We obtained the largest known de-identified dataset from an emotional support system established in Hong Kong, comprising 5682 Cantonese conversations between help-seekers and counselors. Of those, 682 conversations disclosed crisis intentions of suicide. We constructed a suicide-knowledge graph, representing suicide-related domain knowledge as a computer-processible graph. Such knowledge graph was embedded into a deep learning model to improve its ability to identify help-seekers in crisis. As the baseline, a standard NLP model was applied to the same task. 80% of the study samples were randomly sampled to train model parameters. The remaining 20% were used for model validation. Evaluation metrics including precision, recall, and c-statistic were reported. Results: Both KARA and the baseline achieved high precision (0.984 and 0.951, shown in Table 2) and high recall (0.942 and 0.947) towards non-crisis cases. For crisis cases, however, KARA model achieved a much higher recall than the baseline (0.870 vs 0.791). The c-statistics of KARA and the baseline were 0.815 and 0.760, respectively. Conclusion: KARA significantly outperformed standard NLP models, demonstrating good translational value and clinical relevance.
Persistent Identifierhttp://hdl.handle.net/10722/305438
ISSN
2020 Impact Factor: 4.634
2020 SCImago Journal Rankings: 1.913
ISI Accession Number ID

 

DC FieldValueLanguage
dc.contributor.authorXu, Z-
dc.contributor.authorXu, Y-
dc.contributor.authorCheung, F-
dc.contributor.authorCheng, M-
dc.contributor.authorLung, D-
dc.contributor.authorLaw, YW-
dc.contributor.authorChiang, B-
dc.contributor.authorZhang, Q-
dc.contributor.authorYip, PSF-
dc.date.accessioned2021-10-20T10:09:24Z-
dc.date.available2021-10-20T10:09:24Z-
dc.date.issued2021-
dc.identifier.citationSocial Science & Medicine, 2021, v. 283, p. article no. 114176-
dc.identifier.issn0277-9536-
dc.identifier.urihttp://hdl.handle.net/10722/305438-
dc.description.abstractRationale: Detecting users at risk of suicide in text-based counseling services is essential to ensure that at-risk individuals are flagged and prioritized. Objective: The objective of this study is to develop a domain knowledge-aware risk assessment (KARA) model to improve our ability of suicide detection in online counseling systems. Methods: We obtained the largest known de-identified dataset from an emotional support system established in Hong Kong, comprising 5682 Cantonese conversations between help-seekers and counselors. Of those, 682 conversations disclosed crisis intentions of suicide. We constructed a suicide-knowledge graph, representing suicide-related domain knowledge as a computer-processible graph. Such knowledge graph was embedded into a deep learning model to improve its ability to identify help-seekers in crisis. As the baseline, a standard NLP model was applied to the same task. 80% of the study samples were randomly sampled to train model parameters. The remaining 20% were used for model validation. Evaluation metrics including precision, recall, and c-statistic were reported. Results: Both KARA and the baseline achieved high precision (0.984 and 0.951, shown in Table 2) and high recall (0.942 and 0.947) towards non-crisis cases. For crisis cases, however, KARA model achieved a much higher recall than the baseline (0.870 vs 0.791). The c-statistics of KARA and the baseline were 0.815 and 0.760, respectively. Conclusion: KARA significantly outperformed standard NLP models, demonstrating good translational value and clinical relevance.-
dc.languageeng-
dc.publisherPergamon. The Journal's web site is located at http://www.elsevier.com/locate/socscimed-
dc.relation.ispartofSocial Science & Medicine-
dc.subjectOnline counseling services-
dc.subjectSuicide prevention-
dc.subjectNatural language processing-
dc.subjectKnowledge graph-
dc.subjectArtificial intelligence-
dc.titleDetecting suicide risk using knowledge-aware natural language processing and counseling service data-
dc.typeArticle-
dc.identifier.emailXu, Z: zhongzhi@hku.hk-
dc.identifier.emailXu, Y: chicoxyc@hku.hk-
dc.identifier.emailCheung, F: rence@hku.hk-
dc.identifier.emailLung, D: danielwm@hku.hk-
dc.identifier.emailLaw, YW: flawhk@hku.hk-
dc.identifier.emailYip, PSF: sfpyip@hku.hk-
dc.identifier.authorityLaw, YW=rp00561-
dc.identifier.authorityYip, PSF=rp00596-
dc.description.naturelink_to_subscribed_fulltext-
dc.identifier.doi10.1016/j.socscimed.2021.114176-
dc.identifier.pmid34214846-
dc.identifier.scopuseid_2-s2.0-85108917661-
dc.identifier.hkuros328075-
dc.identifier.volume283-
dc.identifier.spagearticle no. 114176-
dc.identifier.epagearticle no. 114176-
dc.identifier.isiWOS:000681172900010-
dc.publisher.placeUnited Kingdom-

Export via OAI-PMH Interface in XML Formats


OR


Export to Other Non-XML Formats