With 1.6 doctors per 1,000 people worldwide and 3.2 doctors per 1,000 people in France, medicine is under immense demographic pressure and providing appropriate access to care to the entire population is still a dream. According to the WHO, 1 in 4 people in Europe are affected by psychiatric disorders during their lifetime. In France, nearly 3 million people are affected by depression each year. An important challenge of artificial intelligence is to facilitate medical diagnoses. Much work has been done in this direction in medical imaging, but other health sectors could benefit from advances in AI. This is particularly the case in psychiatry, where the absence of specific bio-markers or clear imaging results indicating a mental health pathology makes it difficult to establish a diagnosis. Moreover, the diagnosis of psychiatric disorders comes at a high cost because the evaluation sessions are long and require the presence of a qualified professional. In France, the first stages of psychiatric diagnosis are carried out using declarative questionnaires on paper, whose rating is time consuming and prone to errors. These questionnaires assess several dimensions simultaneously, without being made explicit. Reduicing redundancy of the questions and increasing their tailoring to patient’s profile are the challenges that we tackle in this project using advance AI techniques.