Improving OCR Quality in 19th Century Historical Documents Using a Combined Machine Learning Based Approach