際際滷

際際滷Share a Scribd company logo
QA Distiller
               AUTOMATED TRANSLATION QUALITY CONTROL

               Proudly developed by Yamagata Europe




   YAMAGATA EUROPE | Thomas Vackier | March 23, 2011

Wednesday, March 23, 2011
YAMAGATA EUROPE
                                Translating Knowledge




Wednesday, March 23, 2011
What do we do?
               We write

               We translate

               We do the layout

               We print

               We package & ship

               We create software



       Core business
Wednesday, March 23, 2011
On what scale do we do it?
               Team of 50 people

               Pool of 1000 language vendors

               Translate about 400,000 words a day

               Translate into 50 languages

               Layout 450 pages a day

               18,000 publications a day



       Size matters
Wednesday, March 23, 2011
How it all started
                               and got completely out of hand




Wednesday, March 23, 2011
Finding errors manually
               PLAIN BORING

               SUBJECTIVE

               PRONE TO OVERSIGHTS

               NOT EVERYTHING IS CHECKED DUE TO TIME CONSTRAINTS

               LABOUR INTENSIVE




       The European way
Wednesday, March 23, 2011
Finding errors manually
               Our Japanese colleagues were able to find more mistakes

               Even in languages they didn't understand

               Very methodical

               Extremely consistent

               Still very costly




       The Japanese way
Wednesday, March 23, 2011
AUTOMATE
                            What you can objectively measure,
                                   you can automate




Wednesday, March 23, 2011
Enter QA Distiller
               We gain time

               We check everything

               We check objectively

               We check in the same way

               We have reports




       First version developed in 2003
Wednesday, March 23, 2011
What can QA Distiller check?
               OMISSIONS

               INCONSISTENCIES

               FORMATTING

               TERMINOLOGY

               CUSTOM PATTERNS




       Supported checks
Wednesday, March 23, 2011
What cant it check?
               TRANSFER OF MEANING

               GRAMMAR

               STYLE
                                     Human revision stil needed!
               SPELLING*




       Caveat
Wednesday, March 23, 2011
Input formats & languages
               BILINGUAL     TTX (Trados TagEditor)
                             TMX (Translation Memory eXchange)
                             XLIFF (including SDL XLIFF used in SDL Studio)


                                          Bilingual Word not supported
               TERMINOLOGY   DICT (created from Excel using QA Distiller)
                             TBX (TermBase eXchange)

               LANGUAGES     More than 70, including RTL and CJK


       Formats & languages
Wednesday, March 23, 2011
Omissions
               EMPTY                 Spaces or nothing

               FORGOTTEN             Source = Target

               SKIPPED               Untouched

               PARTIALLY FORGOTTEN   Parts left untranslated

               TOO SHORT/LONG        Percentage




       Omissions
Wednesday, March 23, 2011
Inconsistencies
               TARGET LANGUAGE   Switch the unit ON
                                 has been translated differently
                                 1 time(s) as Schakel het toestel aan
                                 1 time(s) as Schakel het toestel uit


               SOURCE LANGUAGE   Schakel het toestel aan
                                 has been used for 2 different source
                                 segments
                                 1 time(s) for Switch the unit ON
                                 1 time(s) for Switch the unit OFF



       Inconsistencies
Wednesday, March 23, 2011
Formatting
               SPACING ISSUES   Double, extra leading, trailing

               PUNCTUATION      Double, end of segment

               BRACKETS         Non-matching pairs

               TABS             Different count

               CAPITALISATION   Initial, entire




       Language independent
Wednesday, March 23, 2011
Formatting
               CORRUPT CHARACTERS   Garbled letters like 鐃,

               SPACING              Required, forbidden

               NUMBERS              Values, formatting

               QUOTATION MARKS      Non-matching, type

               MEASUREMENT SYSTEM   Type, conversion




       Language dependent
Wednesday, March 23, 2011
Terminology
               EXCEL                Convert to DICT using QA Distiller

               OPEN OFFICE          Free plug-in to export to TBX

               SDL MULTITERM 2009   Export to TBX

               TERMWEB              Export to TBX




       Proprietary dictionary and TBX
Wednesday, March 23, 2011
Custom patterns
               SEARCH                Source and/or Target


               REGULAR EXPRESSIONS   Report patterns

                                     Ignore patterns



                                                       Make your own checks!
       Search and Regular Expressions
Wednesday, March 23, 2011
REPORTS
                             Share error logs




Wednesday, March 23, 2011
THE CURSE
                                 of the
                            FALSE POSITIVES
                                and how to avoid it




Wednesday, March 23, 2011
PROFILES
                             Share settings




Wednesday, March 23, 2011
BATCH REPLACE
                               Do more in less time




Wednesday, March 23, 2011
A SAMPLE PROJECT
                               Garbage in, garbage out quality out




Wednesday, March 23, 2011
THE FUTURE
                            The future is bright, and regex-flavoured




Wednesday, March 23, 2011
Get in Touch!

                            thomas@yamagata-europe.com
                                   @QADistiller
                                 @YamagataEurope




Wednesday, March 23, 2011

More Related Content

QA Distiller

  • 1. QA Distiller AUTOMATED TRANSLATION QUALITY CONTROL Proudly developed by Yamagata Europe YAMAGATA EUROPE | Thomas Vackier | March 23, 2011 Wednesday, March 23, 2011
  • 2. YAMAGATA EUROPE Translating Knowledge Wednesday, March 23, 2011
  • 3. What do we do? We write We translate We do the layout We print We package & ship We create software Core business Wednesday, March 23, 2011
  • 4. On what scale do we do it? Team of 50 people Pool of 1000 language vendors Translate about 400,000 words a day Translate into 50 languages Layout 450 pages a day 18,000 publications a day Size matters Wednesday, March 23, 2011
  • 5. How it all started and got completely out of hand Wednesday, March 23, 2011
  • 6. Finding errors manually PLAIN BORING SUBJECTIVE PRONE TO OVERSIGHTS NOT EVERYTHING IS CHECKED DUE TO TIME CONSTRAINTS LABOUR INTENSIVE The European way Wednesday, March 23, 2011
  • 7. Finding errors manually Our Japanese colleagues were able to find more mistakes Even in languages they didn't understand Very methodical Extremely consistent Still very costly The Japanese way Wednesday, March 23, 2011
  • 8. AUTOMATE What you can objectively measure, you can automate Wednesday, March 23, 2011
  • 9. Enter QA Distiller We gain time We check everything We check objectively We check in the same way We have reports First version developed in 2003 Wednesday, March 23, 2011
  • 10. What can QA Distiller check? OMISSIONS INCONSISTENCIES FORMATTING TERMINOLOGY CUSTOM PATTERNS Supported checks Wednesday, March 23, 2011
  • 11. What cant it check? TRANSFER OF MEANING GRAMMAR STYLE Human revision stil needed! SPELLING* Caveat Wednesday, March 23, 2011
  • 12. Input formats & languages BILINGUAL TTX (Trados TagEditor) TMX (Translation Memory eXchange) XLIFF (including SDL XLIFF used in SDL Studio) Bilingual Word not supported TERMINOLOGY DICT (created from Excel using QA Distiller) TBX (TermBase eXchange) LANGUAGES More than 70, including RTL and CJK Formats & languages Wednesday, March 23, 2011
  • 13. Omissions EMPTY Spaces or nothing FORGOTTEN Source = Target SKIPPED Untouched PARTIALLY FORGOTTEN Parts left untranslated TOO SHORT/LONG Percentage Omissions Wednesday, March 23, 2011
  • 14. Inconsistencies TARGET LANGUAGE Switch the unit ON has been translated differently 1 time(s) as Schakel het toestel aan 1 time(s) as Schakel het toestel uit SOURCE LANGUAGE Schakel het toestel aan has been used for 2 different source segments 1 time(s) for Switch the unit ON 1 time(s) for Switch the unit OFF Inconsistencies Wednesday, March 23, 2011
  • 15. Formatting SPACING ISSUES Double, extra leading, trailing PUNCTUATION Double, end of segment BRACKETS Non-matching pairs TABS Different count CAPITALISATION Initial, entire Language independent Wednesday, March 23, 2011
  • 16. Formatting CORRUPT CHARACTERS Garbled letters like 鐃, SPACING Required, forbidden NUMBERS Values, formatting QUOTATION MARKS Non-matching, type MEASUREMENT SYSTEM Type, conversion Language dependent Wednesday, March 23, 2011
  • 17. Terminology EXCEL Convert to DICT using QA Distiller OPEN OFFICE Free plug-in to export to TBX SDL MULTITERM 2009 Export to TBX TERMWEB Export to TBX Proprietary dictionary and TBX Wednesday, March 23, 2011
  • 18. Custom patterns SEARCH Source and/or Target REGULAR EXPRESSIONS Report patterns Ignore patterns Make your own checks! Search and Regular Expressions Wednesday, March 23, 2011
  • 19. REPORTS Share error logs Wednesday, March 23, 2011
  • 20. THE CURSE of the FALSE POSITIVES and how to avoid it Wednesday, March 23, 2011
  • 21. PROFILES Share settings Wednesday, March 23, 2011
  • 22. BATCH REPLACE Do more in less time Wednesday, March 23, 2011
  • 23. A SAMPLE PROJECT Garbage in, garbage out quality out Wednesday, March 23, 2011
  • 24. THE FUTURE The future is bright, and regex-flavoured Wednesday, March 23, 2011
  • 25. Get in Touch! thomas@yamagata-europe.com @QADistiller @YamagataEurope Wednesday, March 23, 2011