A method and a device for detecting plagiarism are provided to detect the plagiarism precisely from a document in a short time without degrading precision by reflecting characteristics of a free word order language such as Hangul or Japanese. A partial similarity of each divided unit is calculated by receiving a document from an input unit, loading the compared document from a storage unit, and dividing the paired documents into each dividing unit(S400). A document similarity between the paired documents is calculated by accumulating the partial similarities when the partial similarity is higher than a predetermined threshold value, and plagiarism is determined on the basis of a total plagiarism level(S500). The partial similarity is calculated by checking a preliminary similarity of a compared part of the paired documents based on a coincident ratio of all divided unit without overlapping, and checking an in-depth similarity in consideration of weight according to a coincident position in the divided unit when the preliminary similarity is higher than the threshold similarity.
展开▼
