When China HowNet checks the duplicate, the pictures and formulas in Word can't be detected at all, while in PDF mode, the pictures and formulas will be detected by processing and identifying them as text, so the identified content is quite different from the pictures and formulas, which is detected as plagiarism, and the PDF is unreasonable.

Moreover, if there are a lot of footnotes, endnotes or contents in the header and footer of the paper, the CNKI duplicate checking system in normal word documents can distinguish these footnotes, endnotes, headers and footers, and these contents are not involved in duplicate checking together with the text. In PDF format, the above parts are likely to be recognized as text, and they will participate in the search together, so the results of repeated search will be different.