What is the best tool to avoid XSLT when writing rules that extract data subset from a large XML-based format into a small Java object?