Sentiment analysis on Turkish Twitter messages by using data mining

One of the aims of this thesis is make a contribution on semantic works for sentiment analysis on Turkish texts. Also, it is called for people who talk in Turkish and having common ground can be brought close together by achieving sentiment analysis in Turkish texts. Thus, social networking sites can be built based on more successful artificial intelligence systems. Tivits of users in a social media, 'Twitter', were analysed as in scope of this thesis. Emotions stated by shared tivits were classified under four main categories. These categories are 'Happiness', 'Anger', 'Sadness', and 'Confusion'. All typo mistakes of tivits were proofread with the help of 'Zemberek' library for classifying these accurately. Proofread tivits were labeled on these four categories by volunteers. After this, these tivits and related results were examined by using the techniques of 'Decision tree' and 'Fuzzy Rules'.