Use tessedit_char_whitelist in Tesseract. It would restrict Tesseract to only the relevant characters which might be helpful for the logographs. Kinda serves a similar role to the dictionary matching but is handled prior to reading rather than in post

[Gin890]