Extraction or recognition
An Assamese PDF saved from a Unicode document holds its text, so converting it is extraction and comes out close to exact. A scanned Assamese PDF holds pictures of pages, so it is recognised first and inherits recognition’s mistakes, described on the Assamese OCR page. You upload both the same way.
Spaces, and the face a scan is set in
A scanned Assamese PDF converted to Word used to come back spaced as English, a short word welded to the next, and set in an English font. It is now spaced as an Indian-language page and set in Noto Sans Bengali, which draws ৰ and ৱ. On our Assamese test page the words matching the reference went from 185 to 203 of 207.
Legacy Assamese fonts
Much older Assamese publishing was typed in Geetanjali and Ramdhenu, legacy fonts that store keys rather than Assamese letters. No table reads them here yet. Bijoy, the Bengali layout we do read, is a different encoding, and reading an Assamese document through it would give wrong letters. So a PDF set in one of these fonts cannot be converted into Assamese letters yet.
What is kept
Assamese text, reading order and paragraphs come through, and so do the fonts a digital PDF names, with their sizes, bold, italic and colour; its tables, as Word tables; its pictures; and its running heads, in Word’s header. A PDF whose every page is scanned is rebuilt a page at a time, at the page’s size, with its lines where the scan sets them. Text wrapped round a picture, or a box laid across columns, comes out simpler than the page.