The rise in mental illnesses and the intricacy of their symptoms make treatment more difficult, which emphasizes how crucial an early and precise diagnosis is. Early detection of mental health disorders is essential to prevent suicide, improve treatment, and promote general well-being. In this context, AI is revolutionizing mental health by enabling more accurate diagnoses, optimized treatments, and better access to care. NLP techniques are also playing a key role in healthcare management. In this work, we use NLP techniques and machine learning models to classify publications related to mental health. Feature engineering is a key element of NLP, transforming raw text into features compatible with machine learning models. In this study, we apply a method based on the TF-IDF technique and two graph embedding methods on a dataset related to mental health sentiments, then train classification models and evaluate their performance.