Setting up a competition framework for the evaluation of structure extraction from OCR-ed books