Development of an AI-Integrated Online Counseling and Self-Improvement Platform for Mental Health Support (Preprint)

<sec> <title>BACKGROUND</title> Mental health issues such as stress, anxiety, and depression are increasingly prevalent worldwide. However, access to professional counseling remains limited, especially in low-resource countries like Sri Lanka. Existing digital mental health tools often rely on internet connectivity and high-end hardware, creating accessibility barriers. </sec> <sec> <title>OBJECTIVE</title> This study aims to design and evaluate an AI-based counseling chatbot that operates offline using a lightweight TF-IDF model, providing accessible, stigma-free mental health support. </sec> <sec> <title>METHODS</title> A dataset of 1,872 mental health-related user expressions was collected from counseling transcripts, affective lexicons, and peer support forums. Text preprocessing included tokenization, stopword removal, and lemmatization. The TF-IDF model was implemented to classify user intents and generate context-appropriate responses from a curated response bank based on cognitive behavioral therapy principles. The model’s performance was evaluated using accuracy, relevance rating, and user feedback. </sec> <sec> <title>RESULTS</title> The chatbot achieved 91.2% accuracy in intent classification, with an average response time of 0.19 seconds and a memory footprint of only 2MB. User evaluation with 10 participants indicated a mean response relevance score of 4.3/5, and 80% reported reduced anxiety after chatbot interaction. </sec> <sec> <title>CONCLUSIONS</title> The proposed TF-IDF-based chatbot demonstrates that lightweight AI systems can provide effective and empathetic mental health support in low-resource environments. Future work includes expanding to multilingual support and integrating with hybrid human-AI counseling frameworks. </sec>