The increasing prevalence of mental health disorders, particularly depression and anxiety, highlights the pressing need for effective early detection mechanisms. Existing approaches often focus on analyzing social media data or utilizing standardized tools like the PHQ-9 and GAD-7. Notably, prior studies on mental illness detection have demonstrated significant success, achieving accuracies of 91% using vector-space word embeddings and 98% when combined with lexicon-based features. However, these methods are limited by their reliance on single-source data, which may not capture the full complexity of mental health states. In this study, we propose a novel hybrid approach that integrates structured data from standardized questionnaires with unstructured data from social media posts. Our methodology employs deep neural networks to process questionnaire responses and Natural Language Processing (NLP) techniques to extract emotional and contextual signals from social media content. This combination enables the extraction of complementary features, enhancing the model’s ability to detect subtle indicators of depression and anxiety. Preliminary experiments demonstrate the effectiveness of our approach, achieving a predictive accuracy of 93%, thus bridging the gap between single-source limitations and comprehensive mental health assessments. This integrated framework opens the door to real-time mental health monitoring and personalized intervention strategies.