From historic books to annotated XML: Building a large multilingual diachronic corpus