While evaluating OCR, some users demand that diacritics should not be considered an error since they are often neglected for information retrieval (search engines). This request raises some doubts about the expected behavior. What do you think about the following issues?