Abstract
The identities of the authors who having phenomenon with their sharings without revealing his/her identity on Twitter which is the most popular microblogging site, are wondered. The sharings of a Twitter account can be used to detect the identity of the user. Especially, a columnist who have written articles on various media organs, even if he/she does not reveal his/her identity, can be guessed. We tried to guess the author of an account by comparing the articles and sharings on Twitter accounts of 10 columnists. We performed tests firstly by taking each tweet as an individual text, and then grouping the specific number of tweets. We perceived that using the grouped tweet texts gives more accurate results than using each tweet individually. Additionally, we caught that we can guess the owner of a Twitter account with a good accuracy rate by comparing the sharings of this account and the articles of the candidate authors. We used the words themselves, their stems and 3-grams for digitizing of the texts. We achieved the most successful results with support vector machines from among several classifiers.