Mental health awareness among students has become a critical concern in recent years, with increasing cases of stress, anxiety, and emotional burnout. Traditional journaling, while therapeutically beneficial, lacks intelligent feedback and emotional pattern recognition. This paper presents DAYRA — an AI-powered mental wellness and productivity web application that integrates a custom emotion weighting and wellness scoring framework built on top of the RoBERTa transformer model. The proposed system detects 28 distinct human emotions from journal text using the SamLowe/roberta-base-goₑmotions model, applies a custom weighted scoring formula to compute a normalized wellness score between 0 and 10, and employs a trend detection algorithm to determine whether a user's emotional state is improving, declining, or stable over a selected time period. The platform further integrates Google Gemini 2. 5 Flash for AI-generated personalized wellness summaries. A full-stack implementation using React. js, FastAPI, and MongoDB Atlas is deployed and evaluated with real users including students, medical professionals, and academicians. User evaluation results demonstrate a 100% satisfaction rate with strong validation from psychiatry and IT professionals. The paper presents the architecture, methodology, formula derivation, implementation details, and evaluation results of the complete system.
Akash Phukan (Tue,) studied this question.