Probably someone entered text encoded in utf8 or another foreign encoding in the fields. If you can read them, it means your editor is configured to do so. Set it to Latin 1 to spot the difference.

You have to identify which encoding was used and hope to code a heuristic to correct those. Like the special start bytes of Utf8 wide characters.

You will end up needing to check and trust the heuristic, or manually correct everything.

As far as I see does ISO Latin 1 not include Greek letters. Hence you have to translate them to HTML entities.

(I suppose that's allowed in the document's definition)