Paperwell Document Text & Data Scan

What it is for

Before you share, redact or cite a document you need two things: its text, and an idea of what sensitive or useful data is in it. This tool does both. It reads the text of a PDF (page by page), a Word file, an HTML page, a text file or a photo of a page, then scans it for:

• identifiers — DOIs, ISBNs, ISSNs, ORCIDs, arXiv ids and PMIDs, with check-digit validation where the format has one;
• personal data to redact — e-mail addresses, IBANs, phone numbers and payment card numbers;
• dates, amounts and periods such as "14 days".

It finds patterns, not meaning: names, postal addresses and ID numbers in free text are not detected. A scanned PDF without a text layer needs OCR first.

Worked example

Paste a short letter:

Dear Ms Doe, your refund of EUR 249.00 will be paid to IBAN DE89 3704 0044 0532 0130 00 within 14 days. Questions: billing@shop.example or +49 30 1234 5678. Berlin, 12.03.2026

The summary: 32 words read (text); 0 identifiers, 3 personal data items and 1 date found. The personal data table lists the IBAN (checksum valid), the e-mail address and the phone number; the date 12.03.2026 becomes 2026-03-12; amount EUR 249.00; period 14 days. Redaction terms are those three values, and the suggested presets are email, phone and iban. "Ms Doe" is not listed — add names yourself.

A 3-page PDF quote with the same kind of data gives 31 words read (pdftotext); 0 identifiers, 3 personal data items and 0 dates found, with 20, 9 and 2 words on the three pages.

Reading notes with doi:10.1287/opre.9.3.383, ISBN 0-306-40615-2, ORCID 0000-0002-1825-0097 and ISBN 978-3-16-148410-5 find 4 identifiers and warn: ISBN 978-3-16-148410-5 has a wrong check digit.

How to read the result

• Metrics count words, pages, identifiers, personal data items and dates.
• Identifiers: as written, normalised, check digit (valid, WRONG, or n/a for DOIs, arXiv ids and PMIDs, which have none) and how often each appears.
• Personal data candidates with their checksum: an IBAN must pass mod-97, a card number the Luhn check, otherwise it is not reported.
• Dates: ambiguous formats such as 03/04/2026 get no ISO date, because the day/month order is unclear.
• Text (if switched on) is the full text, PDF pages separated by form feeds.

Tips and common mistakes

• Upload a file or paste text, not both — the tool refuses mixed input.
• No text from a PDF means it is a scan: run searchable-pdf-ocr first, or upload a photo of the page.
• Text read with OCR gets a warning; check names, numbers and dates against the image.
• A wrong ISBN check digit usually means a typo — verify it before you cite.
• Feed the redaction terms into secure-pdf-redactor, which searches the result again to prove nothing is left.
• Switch off "Include the full text" when you only need the findings; the response is much smaller.

FAQ

Is my file stored?

No. The file is read in memory for this request and not kept. Text input is limited to 400,000 characters, files to 10 MiB.

Why is a name or address not found?

Names and addresses have no fixed pattern, so the scan does not guess. Add them as redaction terms by hand.

Which languages does OCR use?

Photos and scans are read with Tesseract in English. For other languages, make the PDF searchable with searchable-pdf-ocr first.

Does it look up the DOIs and ISBNs?

No, it only finds and validates them. citeshape resolves them into formatted citations.