Narratives of Adolescent Suicide in South Korea: A Text Mining Approach to Online Discussions

Background and Purpose:
Adolescent suicide, as the leading cause of death among youth since 2011, is a significant public health concern in South Korea. Despite efforts to develop preventive strategies, barriers such as stigma, developmental dynamics, and limited access to care persist and further limit accurately identifying and targeting at-risk youth. This study explores how adolescents express suicidal ideation, distress, and help-seeking behavior on anonymous online platforms. We examine online discourse to identify latent themes that capture the lived experiences and psychological states of adolescents experiencing suicidality.

Methods:
We analyzed 6,251 user-generated posts related to adolescent suicide from Naver KnowledgeIN (2014–2022), a widely accessed Korean online Q&A forum that offers complete anonymity and real-time interaction. As one of the most commonly used platforms among Korean adolescents, Naver KnowledgeIN functions as a digital space where youth can disclose emotional struggles freely—often in ways that bypass the formality, stigma, and time lag involved in accessing traditional mental health services. Its structure enables spontaneous, emotionally charged communication in a culturally familiar, linguistically native environment. A four-stage data processing protocol was followed, including data extraction, cleaning, morphological analysis, and topic modeling using Latent Dirichlet Allocation. The final model of thematic clusters was identified through qualitative synthesis.

Results:
Five unique topics emerged from the data corpus of online discussions on adolescent suicide. Family dysfunction topic (24.9% of all data) was represented by family conflict/violence, structural change, and lack of connection--all of which are strained situations that exacerbate adolescents' vulnerability to suicide behavior. The topic of teacher and peer relationships (20.7%) was described as bullying, peer conflict, and negative experiences with teachers. Academic stress and career anxiety topic (12.2%) was associated with exam/grade pressure, academic motivation decline, career uncertainty, and burden of private education. The mental health concerns topic (25.1%) mentioned various forms of psychological distress and negative emotions that are related to suicidal behavior, such as depressive symptoms, emotional numbness, lethargy, anxiety, and internal instability. The help-seeking topic (17.0%) included a spectrum of help-seeking discussions, including indirect pleas for help, urgent appeals for support, and efforts to access professional care amidst structural and interpersonal barriers.

Conclusions and Implications:
This study demonstrates the value of computational text analysis in uncovering mental health narratives within anonymous online discourse, particularly for issues with emotional complexity and cultural uniqueness, such as adolescent suicidality. Notably, anonymous and real-time online platforms may offer immediate and informal venues for emotional expression that may otherwise remain hidden in traditional support systems. This is in contrast with delayed, formal intake channels that are traditionally used for suicide intervention. By treating digital disclosures as informative data on sensitive topics without imposing any additional labeling or stigmatization, we highlight the potential of online platforms to supplement existing mental health services and amplify youth voices in suicide prevention.