QA checking whole TUs in RTF files Thread poster: Ben Jones
|
Ben Jones United Kingdom Local time: 09:20 Japanese to English + ...
Take the following scenario: I have Document A and Document B, with significant repetitions. Doc A was translated as RTF and sent (uncleaned) to the client for feedback on terminology & phrasing, and to a proofreader to spot typos etc. Both client and proofreader marked their changes in 'revision tracking' mode. Meanwhile Doc B has been translated, using & updating the TM generated from Doc A. Now the Doc A revisions have come back, I want to apply them to Doc B. Normal... See more Take the following scenario: I have Document A and Document B, with significant repetitions. Doc A was translated as RTF and sent (uncleaned) to the client for feedback on terminology & phrasing, and to a proofreader to spot typos etc. Both client and proofreader marked their changes in 'revision tracking' mode. Meanwhile Doc B has been translated, using & updating the TM generated from Doc A. Now the Doc A revisions have come back, I want to apply them to Doc B. Normally I should 'accept' the revisions then 'clean' each file to update the TM, right? Then I somehow apply the new TM to Doc B and search for TUs that need to be updated? (Not sure how to do this: set fuzzy matching to 99% and only look for fuzzies?) However... there's a second problem. Imagine Doc A contained: This is a pen. I use pencil. but the client changed this to This is a biro. I use pencil. while at around the same time the proofreader changed it to This is a pen. I use a pencil. and (again at the same time) the translation in Doc B was changed to This is a pen. I use pencils. Trados can't know for sure which version of each TU I want to use in the final TM and final Doc A/B. I'd like to tell it, "Open Doc A/B and display only those TUs where TM_DocB, TM_client and TM_proofreader are different, so I can choose", but I can't see an easy way to do this. Trados does have a "QA Checker" which includes "Repeated segments which have been translated inconsistently" but I'm not sure if that works with multiple TMs, and in any case it can't be used with RTF files, only TagEditor. Any suggestions? [FWIW I'm using SDL Trados 2007 SP2, the original formats were Word, and RTF had to be used instead of TagEditor because of a need to view the formatting while translating.] ▲ Collapse | | |
ApSic XBench | Feb 28, 2008 |
Try XBench. It will let you create a project containing multiple TMs, and you can define which TM is priority. Moreover, you can define one or several TMs as key term files, which means that when you run the QA process in XBench, it will check if the key terms were applied in each segment. You would also need to add the bilingual RTF file to the project and define it as active translation. One way you could use XBench is if you load only one TM at a time in your project and run the ... See more Try XBench. It will let you create a project containing multiple TMs, and you can define which TM is priority. Moreover, you can define one or several TMs as key term files, which means that when you run the QA process in XBench, it will check if the key terms were applied in each segment. You would also need to add the bilingual RTF file to the project and define it as active translation. One way you could use XBench is if you load only one TM at a time in your project and run the QA process. XBench will only propose segments where the key terms TM was not applied consistently. Then, if you right-click on a segment among the results, you get the option "Edit Source", which will open your RTF file with the appropriate segment highlighted (make sure Workbench is running with the appropriate TM loaded). Then, you can edit the segment in Word and save. Once you have edited all segments proposed by XBench, you can run QA again (after you replaced the old RTF file in the project files with the updated one) to see if you've missed anything. If all is OK, then you can remove the TM you used from the project and replace it by the next TM and run QA again. Anyways, XBench being a very complete software for this kind of operation, there are a few other ways you can go about this, still using XBench. XBench is also useful for a bunch of other applications involving TMs and bilingual files. While you're at it, you might as well also download ApSic Comparator - it lets you compare two bilingual files (normally, you compare a bilingual file that was translated, and the same bilingual file in its edited state) and will give you a report of segments that have changed, outlining the changes within segments. Both XBench and Comparator support TTX files, as well as a bunch of other filetypes created with other CAT tools. Give it a try: http://www.apsic.com/en/products_overview.html (download links are on the right) All the best!
[Edited at 2008-02-28 16:12] ▲ Collapse | | |
Trados can't know for sure which version of each TU I want to use in the final TM and final Doc A/B. I'd like to tell it, "Open Doc A/B and display only those TUs where TM_DocB, TM_client and TM_proofreader are different, so I can choose", but I can't see an easy way to do this. Trados does have a "QA Checker" which includes "Repeated segments which have been translated inconsistently" but I'm not sure if that works with multiple TMs, and in any case it can't be used with RTF files, only TagEdit... See more Trados can't know for sure which version of each TU I want to use in the final TM and final Doc A/B. I'd like to tell it, "Open Doc A/B and display only those TUs where TM_DocB, TM_client and TM_proofreader are different, so I can choose", but I can't see an easy way to do this. Trados does have a "QA Checker" which includes "Repeated segments which have been translated inconsistently" but I'm not sure if that works with multiple TMs, and in any case it can't be used with RTF files, only TagEditor. [/quote] I think I would try by creating a new TM, import these three TM into it and set it to display different target text for the same source text (not sure of the exact wording). Then run through your doc B in that, for each segment that is not a 100 % match in all three should show a percentage in the Workbench window underneath the arrows and there you can browse to the different target segment and choose the one you agree with and set that to a 100 % match. But as it is only a theory, don't get to hopeful, but I do a lot of proofreading and I get different targettexts all the time when I work for one particular client who uses several different translators for the same type of texts. I then browse with the arrows and choose the one that I agree with. Regarding tracked changes, yes you accept them and the run translate to fuzzy, I would put 100 % to be on the safe side and set any segments that come up as "yellow" to 100 %. I hope that helps, but I won't guarantee it. Anette ▲ Collapse | | |
Another idea | Feb 28, 2008 |
On second thoughts, maybe we both are complicating things. I take it that the clients changes are LAW and that you can then safely use their TM, and run doc B through it (Doc A's TM) with translate to fuzzy. And then just ignore any changes by the proof reader that are not typos or contradictive of the clients changes. It may sound more time consuming but in the end may turn out to be the more time saving way of doing this job. Anette | |
|
|
Ben Jones United Kingdom Local time: 09:20 Japanese to English + ... TOPIC STARTER
I remember trying ApSic Comparator once in the past, and it was quite handy, so XBench certainly seems worth a look. Anette's method(s) also look promising -- I'll print this out and keep it by the PC while I experiment! FWIW the client's changes are not 'law' in this case: as so often, they know the product/technology well but aren't great on language issues. Many thanks to both of you for your suggestions. | | |