IDML that a CAT tool can read as sentences
If you translate IDML, you know the file a typical PDF conversion makes: a frame per line, a tag per run, and a sentence in five pieces. Ours is built so a segment is a paragraph.
Upload a document6,000 free characters when you sign up. The InDesign file costs nothing on top of the document. See prices See real files
Why it matters
Translation memory matches sentences, and machine translation translates what it is given. A sentence cut in three gets three bad matches, three pieces of nonsense from the engine, and a translator who rejoins them by hand. Tags make it worse: each one costs a keystroke and a QA warning when it moves.
What breaks segments, and what we do
| What breaks | Why | In our file |
|---|---|---|
| Loose frames | A tool reads each frame on its own | Each page's text is threaded through one story |
| Run splits | Every change of run is a tag | Runs that look alike are merged, and a stretch of spaces takes its neighbour's look |
| Fallback fonts | A no-break space in another font shows as a tag inside a word | InDesign's own spaces and hyphens stay in the text's font |
| Soft hyphens | They split a word in two for the tool | They are dropped |
| Hyperlinks | Okapi cuts a link's text out of its sentence | The file holds none |
| Line ends | A PDF breaks lines inside sentences | Such lines run on as one paragraph, with no soft returns in body text |
| Backwards text | Many PDFs hold right-to-left lines as they look | Lines go in the order they are read |
How we check it
Each change to the converter runs through Okapi Tikal, an open source localisation tool that reads IDML with its own filter, and we compare its segments with the source's paragraphs: one segment for each. On our last run across 161 documents in more than 80 languages, all but a handful of lines came out as one segment: a line that split ran a phone number and a web address together or was set in an old custom font.
We have not run these files through Trados Studio, memoQ or Phrase ourselves, so try one document in yours before you commit a project to it. If it behaves differently there, we want to know.
What a translator will still see
- A tag where a run really changes look: a bold word inside a sentence is a tag in any IDML.
- A tag inside some Bengali, Malayalam or Persian words, where the text has a zero-width joiner. Okapi does the same with InDesign's own exports.
- Word comments do not reach the tool: they are InDesign Notes, which Okapi neither extracts nor splits a sentence around.
Trying it in your own tool
- Take a page you know, with a table and a footnote.
- Import the IDML and count segments against the paragraphs on the page.
- Look at the tags: there should be few, each with a reason you can see on the page.
- Pre-translate it from your memory and see how many matches you get, compared with your usual conversion.
- Export the target, open it in InDesign, and look at the Preflight panel.
Language tags in a translation
When you order the translation from us, the whole file is tagged in its target language and no tag of the source stays behind. Left on translated text, a source tag made Word, LibreOffice and InDesign spell-check and hyphenate German as Russian.
Text expansion
A translation usually runs longer than its source. Threaded text carries that on from frame to frame, and when it passes the last frame InDesign marks it as overset instead of losing it. Add a page or a frame, as you would with any file. On a page set as a picture with frames on it, each frame has some room to grow, but check those pages by eye.
A workflow that keeps your tools
- Upload the PDF and download the copy for InDesign.
- Check it in InDesign and fix what you need, so the fixes go into your source.
- Export IDML from InDesign and take it through your CAT tool with your memory and termbase.
- Open the translated IDML in InDesign and finish the layout.
Questions
Does the translated file go back into InDesign?
You can translate the IDML in your CAT tool and open the result in InDesign, or order the translation from us and download it as IDML.
Why not translate inside InDesign?
Because then you lose your translation memory and terminology. IDML lets the CAT tool do its work and InDesign do its.
Are stories on master pages and in tables read too?
Table text is: our check compares every Word paragraph that holds text. How your CAT tool treats master pages is a setting of its IDML filter, so look there.
Try it on a file you would otherwise rebuild
Upload the PDF, the Word file or the scan. Download the copy for InDesign, and the translation too if you ordered one, and open them in your version.
Upload a document