Recently, plagiarism has constantly been in media. With the popularization of the Internet, the increasing availability of ready textual works is notable and the facilities provided by digital media to copy and manipulation of documents, and its contents,. In addition, there is the lack of information and maturity of the students, which have so much information in hand and short time to assimilate it. Students often do not know the plagiarism consequences or even know how to reference properly a work. This dissertation is intended to show different ways of generating and detecting plagiarism, and propose an architecture for a multi-algorithms plagiarism detection tool, that will be used in future researches. The work also presents a validation for the architecture, by implementing a prototype and executing experiments that demonstrate its benefits.