KETCH: A Knowledge-Enhanced Transformer-Based Approach to Suicidal Ideation Detection from Social Media Content

Dongsong Zhang (), Lina Zhou (), Jie Tao (), Tingshao Zhu () and Guodong (Gordon) Gao ()
Additional contact information
Dongsong Zhang: Department of Business Information Systems & Operations Management, Belk College of Business, The University of North Carolina at Charlotte, Charlotte, North Carolina 28223; and The School of Data Science, The University of North Carolina at Charlotte, Charlotte, North Carolina 28223
Lina Zhou: Department of Business Information Systems & Operations Management, Belk College of Business, The University of North Carolina at Charlotte, Charlotte, North Carolina 28223; and The School of Data Science, The University of North Carolina at Charlotte, Charlotte, North Carolina 28223
Jie Tao: Charles F. Dolan School of Business, Fairfield University, Fairfield, Connecticut 06824
Tingshao Zhu: Institute of Psychology, Chinese Academy of Sciences, Beijing 100101, China; and Department of Psychology, University of Chinese Academy of Sciences, Beijing 101408, China
Guodong (Gordon) Gao: Johns Hopkins Carey Business School, Baltimore, Maryland 21202

Information Systems Research, 2025, vol. 36, issue 1, 572-599

Abstract: Suicidal ideation (SI), as a psychiatric emergency, requires immediate assistance and intervention. Most people with SI do not actively seek help from mental health professionals, which may result in irreversible consequences. Research has shown that individuals experiencing SI increasingly express their thoughts and emotions on social media platforms, making the latter a viable venue for suicidal ideation detection (SID). This paper proposes, develops, and evaluates a knowledge-enhanced transformer-based approach (KETCH) to SID from social media content. KETCH comprises several key novel design artifacts, including a social media-oriented SI lexicon, a model-level method for integrating domain knowledge (i.e., lexicon) into a state-of-the-art transformer, and aligned dynamic embedding and lexicon-based enhancement that integrate domain relevance and contextual importance of terms to effective SID. We evaluate KETCH’s performance with social media data in two different languages collected from distinct platforms, and further examine its generalizability to user-level models for suicide risk prediction and depression detection. The results demonstrate the superior effectiveness, robustness, and generalizability of KETCH via a series of empirical evaluation and a field study. Our research makes multifold research contributions and opens up practical opportunities for timely detection and proactive intervention of SI, which can have far-reaching impacts on public health, the economy, and society.

Keywords: suicidal ideation detection; social media; domain knowledge; lexicon; transformer; representation enhancement (search for similar items in EconPapers)
Date: 2025
References: Add references at CitEc
Citations:

Downloads: (external link)
http://dx.doi.org/10.1287/isre.2021.0619 (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:inm:orisre:v:36:y:2025:i:1:p:572-599

Access Statistics for this article

More articles in Information Systems Research from INFORMS Contact information at EDIRC.
Bibliographic data for series maintained by Chris Asher ().