This thesis leverages advanced Natural Language Processing (NLP) techniques to analyze self-reports from adolescents on the Dutch platform DeKindertelefoon, aiming to uncover patterns and themes in their mental health experiences. By employing clustering, topic modeling with BERTopic, and unsupervised classification, the study identifies critical mental health issues such as suicidal ideation, self-harm, anxiety, and substance abuse. The self-reports are classified according to psychological and clinical taxonomies, revealing significant correlations between various mental health concerns. Validation of the model’s performance through expert assessment highlights the challenges and successes in aligning model predictions with expert labels. The findings underscore the complexity of psychological constructs and the potential of NLP methods in mental health research.