Extracting semantic information and structure from documents (VDU).
199 fully annotated forms. Tests semantic entity labeling and linking.