"Probability score": 0.8

"Question improvement suggestion": Please provide more details about the version of Trados Studio you are using and any error messages or issues you have encountered while trying to export in these formats.

In the meantime, based on the information you have provided, here is a suggested answer to your question:

Hello Daniel,

The XLIFF Manager plugin for Trados Studio does support two XLIFF variants, RWS and Polyglott. However, as far as I know, there are no specific XML schema files available for these export formats. The XLIFF Manager uses the standard XLIFF 1.2 and 2.0 schemas, which are widely available online.

If you are encountering issues with exporting in these formats, it would be helpful to have more information about the specific problems you are experiencing.

I hope this answer addressed your question. If not, please reject the answer, add the question improvement suggestion information, and click the RWS Community AI checkbox on the right to try again.

RWS Community AI

Daniel Hug 

The Polyglot XLIFF was created to support the Google Polyglot translation tool.  So we cannot publish the documentation for that.

The RWS XLIFF was our own and it was designed to be simplest and cleanest no frills XLIFF we could manage to work with that we hope could be successfully used in just about any tool. The current version is based on a vanilla XLIFF 1.2, and we had intended to support 2.0 as well, but we never really needed to so it wasn't a priority and we didn't even start it.

Thanks Paul ,

I was looking for a way to hand over a Studio Project to an AI platform where I can work on the translation with AI agents. Then I want to update the Studio files with as little manual copying and pasting as possible. While it might be possible to use the SDLXLIFF file directly, I am afraid of corrupting the file, so I ended up using the XLIFF Manager to export the file, and then hand it to the AI to work on. But I instruct the AI to only return the <target> nodes in order to save output tokens (reduces the output tokens to about 25-30% of what it would need otherwise: <source>, <seg-source>, the definitions etc.).

But that also means I have to merge the AI output with the rest of the XLIFF. I tried to do this with a data mapper which worked like a charm for me until now, but I just could not get this to work – I would have needed a Schema file for the data mapper. The Schema files I produced kind of worked, but never really, yours also did not work for me (sorry to say that).

In the end I resorted to XQuery for the merge, which was simple, but the data mapper would have compiled an executable file. To run an XQuery script I require Saxon, which in turn requires Java, all of which I originally wanted to avoid. Anyway, it's working now, and I just have to stitch it together (automate a few parts of the process) in order to have not a seamless, but a quick and effective integration of agentic AI into my Studio process.

I always liked the XLIFF manager, but I like it even better now. At some point I might have a look at the SDK, it feels like this would be the next step forward for me. But for now, I just needed the added productivity, and it's fit for purpose!

Daniel

Daniel Hug 

Interesting... I did something like that for a customer earlier this week, but not for an AI agent.  I just used Python to build a json file that mapped the content based on the segment ID, did what was needed and then used the mapping file to put it back together again.

It is a lot of fun playing with this stuff.

Paul 

Yes, it's a whole new world. It's like somebody invented the steam engine, and we are now looking where we can make use of all this power that's suddenly available…

Does your method handle tags?

 Daniel Hug 

Daniel Hug said:

Does your method handle tags?

Yes... but it was a different usecase.  I needed to remove trans-units that met a specific criteria so that a cleaner SDLXLIFF could be managed in Trados Studio as there was no way to filter the content using ootb methods.  Once completely translated I then needed to put the trans-units back in again.  So the actual content within the trans-units was irrelevant for me.