Skip to main content
Peer-reviewed Authoritative

Community-level Research on Suicidality Prediction in a Secure Environment: Overview of the CLPsych 2021 Shared Task

Overview of the CLPsych 2021 Shared Task, the first attempt to run community-level mental-health NLP research using sensitive data inside a secure data enclave. Participating teams received access to donated Twitter posts from users with and without suicide attempts and performed all analysis entirely within the secure computational environment, without ever downloading the raw data.

Publisher

Association for Computational Linguistics (CLPsych 2021 Workshop, NAACL)

Published

1 Jun 2021

Added

2 weeks ago

Key Findings

  • Establishes a secure-enclave methodology for shared tasks on sensitive mental-health text, allowing broader community access to suicide-attempt-labeled data without raw data ever leaving a controlled environment
  • Task used Twitter posts donated for research from users with and without documented suicide attempts
  • Reports team results and explicit lessons learned intended to inform future shared tasks on sensitive or confidential data

Methodology Notes

Shared-task overview paper for the 2021 CLPsych workshop at NAACL-HLT 2021 (held virtually, June 2021; exact day not confirmed, day set to 01). Distinct methodology from the already-held 2022 CLPsych task (open Reddit data) — this is the first CLPsych task to use a secure-enclave access model.

Sources

ACL Anthology (primary)

Archived snapshot (Wayback Machine) — preserved against link rot

Authors

Sean MacAvaney, Anjali Mittu, Glen Coppersmith, Jeff Leintz, Philip Resnik

Tags

clpsychsuicide-risksecure-enclaveshared-task2021

Cite This

APA

Sean MacAvaney et al. (2021). Community-level Research on Suicidality Prediction in a Secure Environment: Overview of the CLPsych 2021 Shared Task. Association for Computational Linguistics (CLPsych 2021 Workshop, NAACL). https://aclanthology.org/2021.clpsych-1.7/