I'm downloading post translations using WPML's XLIFF generation functionality. However, I'm finding that the strings/sentences that are part of the same paragraph are grouped together in the XLIFF file, but in the ATE they are separated be default (but you have the option to group them together). Why is this the case? Since the strings are translated separately in ATE it seems like there should be a way to keep them separated when exporting the XLIFF file. Is this possible?
ATE is essentially an external translation service (albeit WPML's own one) which includes some logic itself that is apart from the logic built-in to WPML.
What kind of content is it that you are sending to ATE for translation?
In the case of page-builder content (that includes content made with the core Gutenberg editor, which is handled similarly to other page builders), WPML itself includes logic to break up the content into smaller chunks, e.g. where several different modules are inserted into a page, the content of each of these modules is broken into its own chunk.
So if you were to export an XLIFF file for this content, you would notice it broken up in the same way, because that is handled by WPML itself.
A chunk that was created from, say, a text module that included several paragraphs of text, would still be a single chunk containing all of those paragraphs.
If you were using the classic WordPress editor then the whole of the content, all the paragraphs, would be in a single chunk, as seen when exporting the translation job as an XLIFF file.
Now, ATE includes additional logic whereby when it receives the content (via the XLIFF file) it will further break up the content where it identifies smaller segments of texts (e.g. paragraphs or sentences). So a lump of content from the classic editor would get broken into smaller texts, and—similarly—a chunk of text from a page builder text module would also get broken up into smaller texts.
But when that translated text is ready to be returned by ATE to the site, it assembles the smaller segments back into the same format as the XLIFF file the original content was delivered in.
Back at your site, it receives this XLIFF file but doesn't know anything about the smaller segments ATE worked with and which are now part of the ATE translation memory.
So, to answer your question, no, it is not possible.
Thank you for the detailed response, Nigel. The content we send to ATE is created via the Gutenberg editor in WordPress. The strings that I was specifically referencing are sentences inside of the same core paragraph block. We're interested in having access to the more granular, single sentence/smaller segment strings that are a part of the ATE translation memory because we are exploring a partnership with a translation company that uses Smartcat and we we're thinking forward about exporting/importing our existing post translations into their company's translation memory. It seems more effective to have the smaller segments be a part of the translation memory because there would be a higher likelihood of reuse elsewhere on the website and if we edit that string/sentence later on it would just be affecting the one string instead of the entire paragraph of text that it is a part of.
Yes, so WPML will break up the page into paragraph blocks.
The translation service (in this case, ATE) further breaks the paragraph blocks up into sentences, so that the translation memory is more efficient (it can avoid having to retranslate the same sentence used in different paragraphs in different parts of the site).
So, at the sentence level it is particular to, and confined to, ATE.
Translation memories are not portable across different translation services, implementations differ.
So it is not possible to import the translation memory of another service (e.g. Smartcat) into the WPML translation memory? Not even if it is the same format as the one used by WPML?
It likely wouldn't be the same format used by WPML as every translation service would have a bespoke implementation of translation memory, such that they are not portable.
So, no, importing a translation memory from another service isn't possible I'm afraid.
This ticket is now closed. If you're a WPML client and need related help, please open a new support ticket.
