How to check the translation quality with XBench?

Xbench

 

ApSIC Xbench is an integrated reference tool aimed to provide a clear and structured view of the terminology of any translation project and finding mistranslations, missing translations and tag alterations.

 

Download link: https://www.xbench.net/index.php/download

 

The following are the topics to be treated:

  • How to open a file
  • Different type of Files
  • How to run a QA check and how to select options for a QA Check
  • How to read a report and most common false positives
  • How to save the Report

1) How to open a file

To open a file in Xbench you need to first create a project.

Go to Project, then New (or CTRL + N). A new window will pop up with two tabs. One for Files and one for Settings. The only tab we will be working on is “Files”.

In this tab, click “add” and you will be given all the options of files you can choose from. Once you’ve found the file you will be working with (We usually use TMX and Trados Studio file) click it and then select “NEXT”, you will be taken to the next step where you can “Add files”. Search for the file you want to analyze and then click “Next”. You will be taken to the first screen where you will simply click “OK”.

 

Note: If you’re working with a TMX, you will asked which language is the Target and which one is the Source.

2) Different type of files

ApSIC Xbench allows you to perform powerful searches on the following bilingual formats:

  • Tab-delimited text files (*.txt)
  • XLIFF files (*.xlf, *.xlif, *.xliff)
  • TMX memories (*.tmx)
  • TBX/MARTIF glossaries (*.xml, *tbx, *.mtf)
  • Trados exported memories (*.txt)
  • Trados exported MultiTerm 5 glossaries (*.txt)
  • Trados MultiTerm XML glossaries (*.xml)
  • Trados TagEditor files (*.ttx)
  • Trados Word uncleaned files (*.doc, *.rtf)
  • Trados Studio files (*.sdlxliff, *.sdlproj)
  • SDLX ITD files (*.itd). Note: This option requires that SDLX is installed on the machine.
  • SDLX Memories (*.mdb)
  • STAR Transit 2.6/XV directory tree
  • PO files (*.po)
  • IBM TranslationManager exported dictionaries (*.sgm)
  • IBM TranslationManager installed and exported folders (*.fxp)
  • IBM TranslationManager exported memories (*.exp)
  • OpenTM2 exported dictionaries (*.sgm)
  • OpenTM2 installed and exported folders (*.fxp)
  • OpenTM2 exported memories (*.exp)
  • Wordfast memories (*.txt)
  • Wordfast glossaries (*.txt)
  • Wordfast Pro TXML files
  • DejaVu X/Idiom files (*.wsprj, *.dvprj)
  • DejaVu X/Idiom memories (*.wstm, *.dvmdb)
  • Logoport RTF files (*.rtf)
  • Microsoft software glossaries (*.csv)
  • Mac OS X glossaries (*.ad)
  • Remote Xbench Server glossaries

 

The most common type of files used by our company are TMX memories and Trados files.

Given that MateCat allows you to download a TMX memory, we can run the QA report on it and check for errors, the same happens if we have a TRADOS project.

 

NOTE: all files to be analyzed need to be BILINGUAL that’s why they need to be a memory or a cat tool file such as Trados, if you upload a Word file in English you’ll only get a Spell check.

3) How to run a QA check and how to select options for a QA Check

Once you’ve created your project and opened your file, running the QA check is pretty easy.

The most important part is to have it pre-set correctly.

First thing you need to do is go to the QA Tab where you will find all options.

 

In the Check Group box you will find all categories holding the checklists that can be done.

Please have the following selected:

 

Basic:

 

Untranslated segments

The software will show you all fields which haven’t been translated.

 

Inconsistency in Source

You will be shown translations which are the exact same but their source was different.

 

Inconsistency in Target

You will be shown translations which are different for the same source.

 

Target same as Source

When the target is identical to the source.

 

Content

 

 

Tag Mismatch

The software will show you when tags have been modified

 

Numeric Mismatch

The software will show you when numbers have been modified

 

URL Mismatch

The software will show you when a URL has been modified

 

Unpaired Quotes

The software will show you when quotes are not properly opened and close

 

Double blank

The software will show you when there’s two continuous spaces instead of one

 

Repeated word

The software will show you when the same word is repeated in a sentence

 

Keyterm mismatch

If you’re analyzing a project that has a pre-established memory or glossary, the program will show you any term which is approved already and wasn’t translated as such.

 

Checklist

 

 

Project Checklist

It’s not necessary to check this option.

 

Spellchecking

 

Please select the target language. If you’ve done this with another software, do not select any.

 

Options

 

Do not check any of these options.

 

Once you’ve configured all these checklists, all you’re left to do is click “Check on-going translation”. If you’ve uploaded a partially translated project you will get a partial report while if you’ve already completed 100% of the translation, you should get a final one.

 

If you’ve selected a Spellcheck, that’s the first window that will pop-up and you can go through it like Word’s spellcheck. Once you’ve approved that one, you will get to the report itself.

 

  • How to read a report and most common false positives

The report you will be shown is as follows:

 

 

In “Error Description” you will find the type of error, in this case “Inconsistency in Target”, right next to it you will have the Source and Target columns showing the respective fields.

 

In this case you will see that all source and target texts are the same, yet some are in UPPER CASE, this wouldn’t be an error and this is called a FALSE POSITIVE.

The reason they are called as such is because the program detects that the same words are different and translated differently yet for the final file this is OK as we of course need to respect all upper cases.

 

Most common false positives:

 

  • Upper case not detected and considered a different word.
  • When a number is preferably used written in words in a language the software will detect the number figures are missing.
  • When the author of the text had grammar or spelling mistakes in the source text. Example: We have in one segment “Hi, I’m Gino”, and then in another segment it’s written “Hi, Im Gino” or “Hi, I’m Giino”, the translator will detect the error and translate the same target for both sentences yet the software will tell you the source is different.
  • Target same as Source. Example: Gino is Segment 1, it’s a name so it shouldn’t be translated, Target will be: Gino, yet the software will tell you it wasn’t translated.
  • If you’re checking an unfinished file, the software will show you all blank segments as UNTRANSLATED and MISSED.

 

  • Saving a report

 

To save a report you need to go to QA, then EXPORT QA RESULTS, and you choose the path where you want the file to be saved at (or press CTRL + Q). An HTML file will be saved, it will have the same format as the preview on the software itself.

 

Remember to modify your file accordingly once you get the report and spot any issues in consistency, untranslated segments, double blank spaces, etc. and once you’ve changed them, run the report again to send a clean one to the client.

Leave a comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Stay Update

Be the first to know about our latest editions
and get exlusive invites to our events.

Millennium Translation LLC. © 2017. All rights reserved.