Balanced clustering for efficient detection of scientific plagiarism

Duplication, co-submission and plagiarism are rising phenomena in modern scientific publishing, as the number of peer-reviewed journals and the perceived chances of escaping detection are increasing. On the other side, electronic indexes and new text-searching tools such as the search engine eTBLASTmight provide an effective deterrent of unethical publications. Though manual inspection is unavoidable in the end, automatic detection might strongly reduce the work required. However, the size of online databases makes a full search impractical even by algorithmic tools. In this paper, we consider the problem of structuring a textual database so as to optimize queries for potential duplicates.