Probabilistic-based Text Clustering for Optimizing Mental Health Issues Extraction on Social Media

This study investigated mental health issues in the context of conversations on the social media platform Twitter. Data was collected over the last three years with the keywords mental health, suicide, and depression. The research method uses text mining with the Latent Dirichlet Allocation model. The results of the analysis have been able to identify 8 main topics based on high coherence scores. The dominant topic found on social media Twitter is about religious (20%) and social perspectives on suicide (14,7%). These findings indicate that Twitter conversations often reflect debate and reflection on how religious values and social views affect the perception and treatment of mental health problems, especially suicide. The study provides in-depth insights into the dynamics of online conversations around mental health on social media, highlighting the complexity of the interaction between religious, social, and mental health factors in a digital context. The implications of this study can be used as a basis for the development of more sensitive and effective intervention strategies and communication approaches in overcoming stigmatization and increasing public understanding of mental health issues, especially those related to suicide. The validation of the LDA model for the topic instruction task showed an interpretation accuracy of 83% for word tasks and 87% for topic tasks. This suggests that the LDA model is highly effective in identifying mental health issues, with an accuracy exceeding 80% when aligned with respondents’ views.