Abstract:
"Adolescent suicide is a pressing public health issue, and early detection and intervention are
crucial in preventing such tragic incidents. Although digital interventions in mental health have the potential to increase access to care, reduce stigma associated with seeking help, provide personalized and scalable interventions and improve outcomes for individuals with mental health conditions, existing mental health chatbots lack an effective machine-learning model to accurately detect adolescents' utterances. Additionally, they unable to deliver access to more focused wellness support that teenagers need to receive by utilizing psychiatry domain knowledge in order to prevent teenage suicide attempts by enhancing resiliency.
The Artificial Intelligence Powered voice-enabled chatbot presented in this research paper
represents an innovative approach to leveraging advanced technologies for the prevention of adolescent suicide incidents. The chatbot employs a sophisticated Natural Language
Understanding (NLU) model trained by utilizing different Natural Language Processing (NLP)
techniques and a psychiatry dataset curated from conversations with mental health
professionals and counsel chats. By integrating the chatbot with an upper ontology to leverage psychiatry domain-independent knowledge, the chatbot demonstrates enhanced capabilities in accurately identifying user utterances, supporting with focused wellness support and intervening with adolescents at risk of suicide, ultimately contributing to the advancement of mental health interventions in the digital age.
After performing full NLU evaluation, the sophisticated NLU model abled to achieve 99.4%
Accuracy, 99.4% F1-score and 99.5% Precision, indicating strong performance in accurately
identifying user intents."
