Translate PDF to Uyghur

The Perso-Arabic one, running right to left, which has been the standard inside China since the early 1980s and is what the great majority of Uyghur readers use, including in the Turkish and European diaspora. A Latin alphabet exists and is popular online and in computing, and a Cyrillic alphabet is used by Uyghur communities in Kazakhstan and Kyrgyzstan. If your readers are in Central Asia, or if the output has to pass through a system with weak Arabic support, say so before the job rather than converting afterwards.

Because Uyghur writes all eight of its vowels as letters instead of leaving short vowels off the page, and several of those letters, along with ڭ for ng and a few consonants borrowed from Persian, do not exist in the Arabic alphabet. A typeface built for Arabic simply has no glyph for them. It also cannot join what it does not have, so the missing letters drop out of the cursive line and stand alone, making words look as though they have been broken in the middle.

It is the hardest layout case in this pair, and it is also completely normal for official forms from Xinjiang. The page holds a right to left script, a block of Chinese characters that flows the other way, and Western digits that follow their own rules, and all three have to be positioned under the Unicode bidirectional algorithm rather than dropped into a grid designed for English. Upload the whole page rather than cropping to one language, because the layout is part of what has to be reproduced.

Keep the differences visible rather than tidying them away. A Uyghur name can appear in Arabic script on one document, as a pinyin transliteration through Chinese on an identity card, and in yet another Latin spelling on a travel document, and none of the three matches the others. In an immigration or asylum file, an unexplained mismatch is a problem, but a silently harmonised one is worse. The usual practice is to render each document as it stands and explain the relationship separately.

A great deal, because most of what arrives in this pair is photographed rather than exported: a page held under a phone camera, a certificate copied several times, an old document that has been folded. The reading step, not the translation step, is where quality is lost, and no amount of language modelling recovers characters that were never legible. If a sharper copy of a document exists, it is worth waiting for.

Use it to work out what you are holding, particularly when a file arrives as dozens of photographs and someone has to triage it quickly. Do not file it. Protection claims are examined closely, and dates, place names, institution names and the spelling of personal names are exactly the fields where an automated reading of a poor scan goes wrong. The submission needs a certified translation signed by a qualified translator.

Yes, and it is the more common direction by a wide margin. Lawyers, caseworkers, researchers and journalists need to read identity documents, transcripts, medical records, correspondence and testimony that exist only in Uyghur. The reverse direction comes up mainly for organisations producing information for Uyghur communities abroad, where the Arabic alphabet is again the right choice for most readers.