Traditional methods of mental health assessment, such as periodic surveys and face-to-face consultations, often fail to capture the dynamic and fluctuating nature of an individual’s emotional state. Mental health issues are increasingly prevalent, yet early detection and continuous monitoring remain significant challenges due to barriers in accessibility, stigma, and limited clinical resources. This study presents a deep learning framework designed to analyze text-based inputs for accurate, continuous emotion detection, and mental health assessment. Leveraging natural language processing (NLP) and advanced transformer-based models, the framework quantifies emotional expressions along a multi-axis scale, capturing nuanced shifts in user emotions over time. This system processes daily text entries to detect patterns indicative of psychological distress, such as anxiety, frustration, or apprehension, and can identify trends that may signal the need for professional intervention. The model integrates BERT transformers for high-fidelity language understanding, paired with recurrent networks (LSTMs) for sequential emotion tracking. An emotion axis system further enhances the model’s capacity to map emotions on a scale, enabling more precise and personalized insights. This approach supports both self-monitoring and early mental health screening, offering users tailored recommendations for coping and recovery while ensuring data privacy and ethical use. The Model demonstrated a positive impact by accurately detecting emotions with high confidence (e.g., joy 0.89, excitement 0.03, gratitude 0.01), processing 1.616 samples/sec, and outperforming VADER and TextBlob in delivering detailed emotional insights, enhancing mental health assessments.