An Improved Cosine Similarity Algorithm Based on Document Similarity

With the coming of big date and high development of education level, there are more and more duplicate documents, which does not only waste source but also time of readers. Therefore, duplicate checking is extremely important. In order to achieve better similarity measure result, we usually transfer documents into distance, angle or curvature. Cosine Similarity is a common method used to measure the similarity of documents, but it is insensitive to number and proportion, which brings some problems to duplicate checking. In this paper, we put forward a new similarity measure method and add sliding window in Cosine Similarity so as to improve its efficiency. With repeated experiments, we finally found that the improved method not only overcomes the drawbacks of Cosine Similarity, but also is superior on speed and accuracy.