Workflow with lots of repetitions

Marie-Claude Falardeau wrote:

I usually merge the files in 1 file when creating the project in Studio, and make sure auto-propagation to confirmed segments is activated. This way if I make a change in one of the reps, confirming the segment will propagate the change to all the other segments.

Giovanni Guarnieri MITI, MIL wrote:

send him the merged sdlxliff file... or maybe he hasn't got Studio?

Heinrich Pesch wrote:



Even if one reads the xliff-file, you never know if it the segment aupdated and ok or a rep from the first translation.
But mostly proofreaders read in Word.

Shai Nave wrote:

Under Project Settings > All Language Pairs > Batch Processing > Analyze Files > Export Frequent Segments.
This will create a SDLXLIFF file of all the frequent (based on the value you define) segments. Translate this file and send it for proofreading (by using the export for review feature). When it is returned, update the translation (using the update from external review feature) with any changes that were made by the reviewer.
Then proceed to update the TM with these repeated segments, pre-translate the files and set to lock these segments. Filter for unlocked files in Studio's Editor view and translate the rest of the text.

The only caveat is that the reviewer needs to use Studio too (or other Translation Environment Tool with good compatibility to Studio) in order to take advantage of this workflow. I don't think that it is possible to filter the segments that will be exported using the Export for external review feature.

Heinrich Pesch wrote:



Exporting unknown segments should deliver a file with all segments that are no 100% matches and one instance per segment.
But it does not seem to work, I had merged 82 files and the export-function just does not do anything.

Rudolf Vedo CT wrote:

Are repetitions even a meaningful concept when it comes to proofreading? I've done many proofeading jobs in my career but have never been asked to disregard or discount repetitions when proofreading (except perhaps once, by a slightly clueless new PM whom I promptly educated), even for jobs with many repetitions and/or when the proofing is done in a CAT tool. Even though a tool has an autopropagation function, the autopropagated text can still be incorrect at any location, be it for reasons of context, computer glitch or incorrect settings.

Heinrich Pesch wrote:

How to avoid reading all those reps when proofreading? How to avoid paying your proofreader for reps?

Jo Macdonald wrote:



Hi Heinrich,
If there are say 80 thou reps and the proofreader is only going to see one segment of those 80 thou as one of 20 thou others, I don't think that's a good idea. When proofing a job you want to see it as a whole, finished work and not as parts.
Say there was a minor glitch in that one segment the proofreader checked and they didn't catch it or consider it to be that important as one of 20 thou others? When it's repeated 80 thousand times it might become much more relevant, and even be out of context in some cases.

In terms of pricing I'd say most people would accept a discount for such a repetitive job.

On the question of the auto-propagate feature in Studio, imo it can be a real waste of time and lead to pasting errors into segments that are properly translated so you need to go back and change them again. Personally I never use it and consider it to be almost as bad as a MT function.