I'm reading the segmentation documentation. I'm missing something - it seems to suggest that segmentation rules are in TMs. But before one ever creates a TM, one begins translating in Studio. And it's breaking the original text into segments. For literary translation I want the segments to be smaller - much smaller. Not sentences, but phrases. In short, what am I missing? Thanks in advance.
Hi John,
When you begin translating in Studio, you will probably have created your very first TM already. Any single file opened will follow the default rules of the TM that appears at the top of the list under General Options. (File>Options>Language Pairs>All language pairs>Tm). Click the TM at the top of the list (or the only one there), then click Settings>Language Resources>Segmentation Rules>Edit.
Here, you can select Paragraph based segmentation. Perhaps this is what you're looking for? Studio will start a new segment every time it comes to a hard carriage return.
After changing any default segmentation rules, you will need to process your file again for the new rules to be applied.
HTH,
Emma
Paul Filkin | RWS Group
________________________
Design your own training!
You've done the courses and still need to go a little further, or still not clear?
Tell us what you need in our Community Solutions Hub
Unknown said:if I wanted to use paragraph segmentation for literary translation and sentence segmentation for legal translation, is there any easy way to shift between them? What are the options that come to mind?
Two separate translation memories is the way to go. One with paragraph segmentation and the other with sentence segmentation.
Then, to organise your work even better, set up two separate projects, one literary, one legal, and simply add new files to those projects as you go.
If, for some reason you want to, say, refer to your legal TM when you're doing a literary translation, simply add the legal TM as the 2nd in the list in that project. It won't affect text segmentation but you'll be able to use it for reference (concordance).
Unknown said:I see, that makes sense. FYI, I'm looking for segmentation by the smallest possible unit. The units of translation, functionally, are words and phrases. In truth, I find the basic principle of choosing longer segments surprising. What I mean is: the program specifically defaults to NOT breaking at a semicolon - and yet you should have complete grammatical sentences before and after a semicolon. Why would you want to establish a standard only for this COMBINATION of sentences, and not for each sentence in the combination?
What use would a TM containing such (long) segments be to anyone?
Ideally, I'd be able to have two entries - the one that'll be expressed in "Save target as" and a (potentially very different) one that would go into the TM.
Unknown said:One more question that occurs to me - if I wanted to use paragraph segmentation for literary translation and sentence segmentation for legal translation, is there any easy way to shift between them?
Surely you'd use two TMs in this case? I doubt you'll get much leverage for a literary translation out of a legal TM anyway, and vice versa. If you don't then perhaps you should.
And the easiest way to manage this is to use Project Templates... one for legal and one for literary work. Then creating projects is simple as the templates are already set up to reflect the type of translation it is.
Paul Filkin | RWS Group
________________________
Design your own training!
You've done the courses and still need to go a little further, or still not clear?
Tell us what you need in our Community Solutions Hub
Unknown said:FYI, I'm looking for segmentation by the smallest possible unit.
Can you precisely define the term "smallest possible unit" using "TECHNICALLY" RECOGNIZABLE terms like number of characters, kind of characters, sequence of certain characters, etc.?! Simply something TECHNICALLY analyzable/recognizable by the segmentation engine... I doubt so. And that's exactly the reason why it cannot be done like you want and why it's done in CAT tools like it is done.
What you are asking for is some AI superengine which would actually UNDERSTAND THE MEANING of the text and divide it to LINGUISTICALLY "logical" parts (segments). But then you are in a completely wrong forum... you should perhaps check some Google or MIT labs or similar...
Translate
