• State Verified Answer
  • Replies 3 replies
  • Subscribers 433 subscribers
  • Views 1323 views
  • Users 0 members are here
Options
Related

Best process for creating a package and other project management hints.

Keenan Cooper

Hello,

I'm looking for advice particularly from those who work for translation companies ( or those who freelance.

I'm creating a translation package for another department in my company sent out to be translated. Working from Japanese to English. We use Studio 2017. Don't know version the translation company uses. Let me know if I'm doing anything wrong or unnecessary. Or, if there's anything I could do that would make the translator's job easier, or save us money/improve results.

1. Create termbase for project
2. Cleaned up segments (removed line breaks within original document)
3. Ran it through a basic TM to see if I could tinker with any segmentation rules for better results
4. Created a project with an empty TM for updating and our company's TMs for resources.
5. Added the TB I made for the project
6. Added an autosuggest dictionary based on our TMs (is this necessary/useful at all??)
7. In editor view, scanned the document for any stray broken segments and combined them.
8. Use the filter and a regex to filter out any segments that contained no source language characters, copied source to target, and confirmed segments
9. Ran pre-translate using our resources. Set the match accuracy level to 80% (should this be lower? Higher? This is the value I normally use). 100% and CMs set to confirm. (this stage now)
10. Run analysis at this state (should the analysis be done before? does it matter there have been some translations added?)
11. Save the analysis as an excel, shop around to a few translation companies, and we'll pick the best deal. 
12. From here, take the project and its resources, turn it into a package, chuck the package into dropbox (file size limit excess e-mail) and send the link to the translation company.

Thoughts? I need to make a guide for my coworkers so they can start making packages as well. 

I'm curious about what kind of settings you (as the customer) use when sending files out for translation. Does anyone use the QA settings to throw an error if a term is not properly used? 
Just starting out with project management so any real advice based on experience would be appreciated.  
Rate translation
×
Suggest better translation
×
Moderator UI
×
  • Get AI Suggestion
Parents
  • 0

    Hi Keenan,

    Here are my 2 cents:

    > 6. Added an autosuggest dictionary based on our TMs (is this necessary/useful at all??)
    If you are using the Fragment Match feature, it pretty much eliminates the need for autosuggest.
    Otherwise, autosuggest can be helpful.

    >7. In editor view, scanned the document for any stray broken segments and combined them.
    You could just leave this up to the translator.

    > Set the match accuracy level to 80% (should this be lower? Higher? This is the value I normally use). 100% and CMs set to confirm. (this stage now)
    I wouldn't change the accuracy level as the translator will see the fuzzy match anyways when they select the segment.

    >8. Use the filter and a regex to filter out any segments that contained no source language characters, copied source to target, and confirmed segments
    I would just lock these segments rather than copying and confirming them.

    >10. Run analysis at this state (should the analysis be done before? does it matter there have been some translations added?)
    Not sure what you meant by "does it matter there have been some translations added". Who is adding the translations?
    Analysis at this state is correct as you need to factor for what you did at step 8.

    >Does anyone use the QA settings to throw an error if a term is not properly used?
    We use specially configured QA settings and ask the translator to run it after finishing the project (along with a spell check)
  • 0 in reply to Jesse Good

    Hey Jesse,

    Appreciate the input. My responses are in blue.

    > 6. Added an autosuggest dictionary based on our TMs (is this necessary/useful at all??)
    If you are using the Fragment Match feature, it pretty much eliminates the need for autosuggest.
    Otherwise, autosuggest can be helpful.
    Got it. I'll include it for now as I'm not sure what if the companies we deal with use Studio 2017.

    >7. In editor view, scanned the document for any stray broken segments and combined them.
    You could just leave this up to the translator.
    I've asked someone else about this in the community and they basically said that a translator will often just translate a document as is. So far for results back from some companies, I've seen poor practices like a split Japanese sentence translated with the English parts in the non-corresponding segments (to maintain natural English) or with parts in the corresponding segments but end up sacrificing the English grammar. If not that then giant block segments from table of contents sections where the translator didn't split things along soft breaks. So I've taken it upon myself to clean things up first as well as nail down the segmentation rules. Maybe it depends on the translator to take the initiative and fix things.

    > Set the match accuracy level to 80% (should this be lower? Higher? This is the value I normally use). 100% and CMs set to confirm. (this stage now)
    I wouldn't change the accuracy level as the translator will see the fuzzy match anyways when they select the segment.
    So pre-translating the translation using segments over 80% doesn't really make a difference? Okay. 100% then.


    >8. Use the filter and a regex to filter out any segments that contained no source language characters, copied source to target, and confirmed segments
    I would just lock these segments rather than copying and confirming them.
    I'm assuming that on a bill/estimate that locked segments are handled as CM and therefore aren't charged, eh? I'll try that next time and see what the estimate looks like.

    >10. Run analysis at this state (should the analysis be done before? does it matter there have been some translations added?)
    Not sure what you meant by "does it matter there have been some translations added". Who is adding the translations?
    Analysis at this state is correct as you need to factor for what you did at step 8.
    Translations are added by me from pre-translate after I've run the regex for no source language characters. I previously did the pre-translate with 80% but now I'll just do 100%. The last project had a status of 10% complete after the regex/pre-translate processes. THEN I did the analysis. So the analysis done at this stage is correct.

    >Does anyone use the QA settings to throw an error if a term is not properly used? 
    We use specially configured QA settings and ask the translator to run it after finishing the project (along with a spell check)
    Got it. Once I get the hang of things I'll take a swing at customizing the settings. 
  • 0 in reply to Keenan Cooper
    Hi Keenan,

    > Maybe it depends on the translator to take the initiative and fix things.
    Yes, also sometimes you may not want the translator to modify the source, so there could be times when you do not want them to fix it for you. However, it is a difficult task I think to fix everything up front especially if you have very large documents, so I think it is more efficient to have the translator do it as she translates.

    > I'm assuming that on a bill/estimate that locked segments are handled as CM and therefore aren't charged, eh?
    Yes, or simply excluded from the analysis.
Reply Children
No Data