Opinion extraction and sentiment detection for turkish documents using machine learning techniques

This study examines the relationship between the use of Turkish and psychological states. The Turkish texts used were collected from depression, mania and healthy people in adults, depression and Attention Deficit Hyperactivity Disorder (ADHD) in children and healthy people. In analyzing the texts, various feature extraction methods have been tried to reach the best result. Feature extraction methods most commonly consist of two types. List method and Single method. In the List method, distinctive words/categories are collected in the lists. In the Single method, words/categories are taken as individual features. The sub-methods consist of 8 lists, 4 singles and 8 mixed methods. Features are obtained using the results of morphological analyzes. Morphological analysis of texts is done by morphological analysis program. In another program called Weka, analysis is done by classification methods using Machine Learning techniques. An application has been developed to identify differences in the word usage of diagnosed persons. The results of this application show that the use of vocabulary in Turkish gives clues about the psychological state of people. This study is the first step of an application program for the psychiatrist.