Abstract:
The accuracy and effectiveness of mental health screening systems depend largely on the quality
of the data and the models used to interpret user inputs. Traditional screening methods, such as
fixed questionnaires, often lack emotional depth and fail to provide personalized assessments. This
creates a significant challenge, especially when detecting depression through open-ended
responses, as understanding emotional context and sentiment in text requires robust Natural
Language Processing (NLP) and machine learning techniques.
This study presents MindEase, a depression screening web application that combines structured
questionnaires with free-text responses analyzed through sentiment classification and a
conversational chatbot. To achieve accurate mental health assessments, the system uses machine
learning models trained on publicly available datasets, enhanced by TF-IDF feature extraction and
sentiment scoring through a logistic regression classifier. Furthermore, the system integrates the
Gemini API to dynamically generate personalized follow-up questions, improving the engagement
and depth of user interactions.
Experimental results show that the implemented model achieves a high level of accuracy in
classifying sentiment, aiding in detecting potential signs of depression. By combining structured
and unstructured data analysis with real-time feedback, MindEase offers a more adaptive, user
friendly, and emotionally aware mental health support tool. The system also adheres to privacy
standards and aims to lower the barrier for individuals seeking early mental health assistance.
