Non-relevant document reduction in anti-plagiarism using asymmetric similarity and AVL tree index

Anti-plagiarism applications have been developed using various approaches. Many methods compare one document to others, regardless of their relevance. This paper proposes a method to reduce non-relevant documents (those having no similar topic with query document) by using asymmetric similarity. Whole documents are collected in one corpus. Each document is preprocessed using winnowing algorithm. The feature from winnowing is then indexed using AVL Tree algorithm to fasten document comparing process. The result shows that reducing non-relevant document shortens almost 10 times of the processing time compared to non-reduced process. Meanwhile, both processes show the same accuracy of 89.78% to give suspected documents.