CafeTran

Hey Igor,

I've finally got the Preview via LO working and I kind of like it. I can see why you went this way, because of cross-platform compatibility and return on invested development.

I can also see Mario's problem with repetitive segments, that even when taking the context into account cannot be identified 100 % reliably.

How about this approach? Add two keyboard shortcuts or buttons to either CafeTran Espresso or a floating pallet that allow the user to navigate to the next/previous occurrence of the segment? Some manual labour of the translator shouldn't be insurmountable.

Cheers,

Hans

Hey Mario,

>The point is that, especially with technical documents, being bounced back to the first instance of those segments is a nuisance

Why do you say that? Is this really a problem, apart from the 'flashing'? Because: as soon as a unique segment after this repetitive one is detected, you're back on track.

Hans wrote: How likely is it that two or more identical segments have the same context and yet require different translations?

The point is that, especially with technical documents, being bounced back to the first instance of those segments is a nuisance and makes no sense from the translation ergonomics viewpoint. At that point, I would much prefer the in-tab HTML preview if CTE decided to make the conversion automatically.

Yep, less keyboard shortcuts/buttons is more too!

However, it's like ChatGPT's 'smart' answers; automation is nice, but some user interaction will be necessary.

I haven't tested the context-sensitivity of the preview function yet, but I was wondering:

How likely is it that two or more identical segments have the same context and yet require different translations?

Hans wrote: Add two keyboard shortcuts or buttons to either CafeTran Espresso or a floating pallet that allow the user to navigate to the next/previous occurrence of the segment?

This might be an idea, although adding yet another click to the translation process wouldn't get me very excited.

I'm don't know how CTE is currently managing the real-time preview, and I'm very far from being able to suggest a possible solution. I only thought that it might be possible to investigate whether a sort of match can be attempted between LO's internal line numbers and CTE's internal segment numbers (of course provided that such numbers actually exist and can be matched), and whether CTE can be told to go ahead only while translating.