Irony detection on Turkish microblog texts

Taşlıoğlu, Hande
Social media is the new trend for expressing personal ideas to other people. Since people are sharing real time messages about their opinions on diverse topics, there exists huge amount of raw data to analyze. Thus, manual classification of these data becomes impossible. Irony, as a simple definition, is creative use of language and attracts computer scientists’ attention lately. Automatic detection of irony on microblog texts is not a trivial task. Texts of microblogs can have limited number of characters, mostly include typing errors therefore traditional methods of opinion classification cannot be applied easily. Therefore, a preprocessing requirement is occurred. After preprocessing, some patterns and language specific features are extracted in order to detect irony. This study aims to automatically detect the irony in microblogs, i.e., informal short texts. Due to the morphological structure of Turkish, various methods are applied to increase the success and quality of classification.