In this research, language analysis of texts is carried out in order to determine whether subjects are healthy or have a psychiatric disorder. The branch of Natural Language Processing (NLP) applied in this research is opinion mining and sentiment analysis (OMSA). Texts are collected from both psychiatry patients and normal controls. First, the most frequently used words for each group are identified. Then, the frequency of these words are used by each person is measured. Using Na ve Bayes Algorithm as a Machine Learning (ML) technique, this data is then used to train the system. The most accurate technique is found to be Na ve Bayes and the results are provided.