The project aims to improve the search on a Nineteenth-Century American Newspapers corpus using and developing data mining tools. It is working spaces-efficient n-gram indexing to identify candidate newspapers and then exploits local models of alignment to identify reprinted fragments unknown a priori.