Six semester student at the gunadarma university obliged to make the writing of scientific or scholarly research, enabling students to take writing materials owned by others. to overcome these problems, it is not enough simply to remind the students that such action is not good. detection of text document similarity is one solution that should be done so that fraudulent activity can be minimized. systems or tools to detect plagiarism is quite a lot, both to detect the text or document or to detect the source code programming. systems or tools to detect the similarity of documents in english have been widely developed. examples of tools that have been developed for english-language document are turnitin, eve2, CopyCathGload, WoedCheck, Glatt, moss and Jplang and so forth. similarity detection research done on text documents or documents written in indonesian text was still relatively little done. in this paper will present an analysis of several methods to detect similarities documents written in indonesian. the methods used are the keyword method, the karp and rabin method and jaro winkler distance method. data used for this test is abstraction of scientific writing data gunadarma university information systems majors. the data abstraction will be modified into three kinds of abstraction: first, there was only a partial abstraction of the same data abstraction will be compared. second is the abstractions that sentence changed positions from data abstraction are compard and the third is abstraction whose content is replaced with a synonym of the word in the abstract that will be compared. from the test results, the best method is the method of rabin karp, except for a synonymous.