AnAbsolute Development of a Clinically-Balanced Digital Journaling System Using Artificial Intelligence for Mental Health Monitoring by using Sentiment based Feedback
Year: 2025
Type: dissertation
Abstract: The growing global burden of mental illness, compounded by long waiting lists and the under-diagnosis resulting from reliance on retrospective self-reporting, highlights an urgent need for scalable and continuous mental health monitoring solutions. Current digital health tools often suffer from low user engagement and lack the contextual depth necessary for clinical insight. This thesis aimed to develop and evaluate a novel, Clinically-Balanced Digital Journaling System augmented with Artificial Intelligence. The primary objectives were to: (1) accurately generate emotional insights from user text entries via sentiment analysis; (2) serve as an early warning detection tool for potential mental health decline; and (3) provide personalized, reflective guidance to promote proactive self-care. The system features a React-based front-end and a Python Flask API back-end housing a robust Transformer-based (BERT) model. This model was fine-tuned on the GoEmotions and Sentiment140 datasets (Demszky et al., 2020; Go et al., 2009) to classify fine-grained emotions and sentiment polarity in diary-style text. Safety was paramount, utilizing a controlled, rule-based approach for generating empathetic, supportive feedback and a hybrid classifier-keyword filter for crisis detection. Through scenario-based evaluation on simulated journal entries, the AI pipeline demonstrated strong performance, achieving an overall emotion classification accuracy of approximately 88% on validation data. Critically, the crisis detection module achieved a 100% detection rate for controlled critical cases, prioritizing user safety (high recall). The system delivered near real-time feedback, with 90% of entries processed in under one second, affirming its practical feasibility and high usability. The project successfully demonstrated the feasibility of an AI-powered journaling system that merges state-of-the-art Natural Language Processing (NLP) with an ethical, user-centric design. This system functions effectively as an interactive mirror for the user’s emotional state, laying critical groundwork for the next generation of scalable mental health self-monitoring tools that complement, rather than replace, professional clinical care.
Source: Theseus (Ammattikorkeakoulujen)
Author: Islam, Nahid
Institution: JAMK University of Applied Sciences
Language: English
Topic: Mental Health via Writing
Subfield: Social Psychology
Field: Psychology
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7115821596
