A localization engineer is both an IT expert and a translation expert who posseses advanced knowledge and skills from both fields. The tasks of a localization engineer have evolved with changing technology so that translators are not burdened by technical problems.

The starting point for successful localization engineering of a localization project is the localization workflow. Larger chunks of work need to be broken down into smaller manageable tasks. These tasks need to be clearly defined where they begin and end so that the localization workflow can be seamlessly intragrated into a larger workflow.

Services provided by a localization engineer typically include any combination of the following tasks in the localization workflow:

  1. Define scope
  2. Define functional requirements
  3. Define quality requirements
  4. Select CAT tools
  5. Prepare translation memories
  6. Prepare translation files
  7. Prepare terminology files
  8. Perform translation
  9. Post-edit
  10. Retrofit translation files
  11. Review
  12. Cleanup

Define scope

Taking the time to define the appropriate scope of you localization project is important so that you do not waste resources, spend to much time/money or have to make changes in the middle of the localization project.

Questions to consider:

  • Where does the project begin and end?
  • How do you know the project is successful?
  • Will the results of the localization project be used only one time, or should they be reused?
  • Do you need to reuse or modify translations for the future?
  • Do you have resources (personnel, infrastructure, time, money) available for setup and maintenance?
  • Have you consulted with your IT department about security?
  • Do you want to use cloud services or on-premise solutions?
  • Does your localization workflow need to be integrated into a larger workflow (example: Scrum or Kanban)?
  • Do you need custom scripting or coding (example: PHP, Python, Linux bash, or Windows PowerScript)?

Define functional requirements

This can be use case scenarios to describe how the localization should fit into your existing workflow.

Questions to consider:

  • Who will be working with the localization?
  • What do they need to achieve?
  • Are there preexisting technical requirements?

Prepare your style guide

Select locale

Define the most important linguistic aspects


Define quality requirements

Defining a clear and simple quality definition will help you later to make quick and informed decisions about the tools, technologies, and options which you will use in your project.

The project management triangle describes quality as a function of time, cost, and scope. Of course, we always strive for maximum scope. However, the scope of your project may be restricted by cost or time right from the start. The first step is to make a list of quality aspects which are important to you and prioritize them.

Questions to consider:

  • Which translation errors or formatting errors may occur? (error allowance)
  • How severe are the errors which may occur? (error tolerance)
  • How often may the errors occur? (error frequency)

Linguistic items may include the consistent usage of specialized terminology, company style, locale, and sentence structure.

Other considerations may include how Unicode characters, HTML entities, or inline tags in XML are handled during translation.


Select CAT tools

(CAT = computer-aided translation)

Now you are ready to collect your tools and materials for translation.

Questions to consider: 

  • Do you already have translation software which you have to use?
  • Does your text have a lot of tags?
  • Do you need to reuse or edit translations?

If you need to reuse or edit translations, then you need a translation memory. There are many products on the market to to manage translations. Alternatively, custom scripts can also be written to store the translations to a database.

Also you need to leverage this data before translation begins (pre-translation). CAT tools provide comprehensive settings to complete this as much as possible.

When pre-translation is done, machine translation can be applied. In the CAT tool, an API can be specified for a machine translator. Then only text which is not pre-translated is processed.

Neural MT

Does the machine translator support your required locale?

Does the machine translator support different styles?


Prepare translation memories

Do you have existing documents which can be added to a translation memory?

Clean data

Do you have an existing translation memory?

Should only some of the translations be used?

Does your existing translation memory need to be converted (example: xlifv1, xlifv2)?

Do you need to reverse the source and target languages (locales)?

Does your translation memory need to be cleaned?

Do new translation memories need to be created?

How should new translation memories be configured?


Prepare translation files

Does the source text need to be extracted from code?

Are the source text files in a mark-down format?

Do you need a special filter for your file type?

Do you need to convert files (example: reST to DOCX, Pandoc, Sphinx, DocUtils, rst2doc)?


Prepare terminology files

Do you have specific terminology which needs to be applied?

Then you need to make sure that the machine translator supports this functionality and the languages and locales that you need.

Do you need to convert your translation memory (example: CSV, TBX, or XML)?

Do you need to extract your terminology from a database (example: MySQL)?

Do you need to use your translation memory as a source for terminology?

Do your terminology files need special preparation for the machine translator?


Perform translation

One of the most overlooked steps in machine translation is preparation and cleaning.

Perform translation

Post-editing

Human review with style guide

AI review?

 

What about "human translation"?

It still exists, and there are certain advantages, but all in all it takes more time.

 

Privacy considerations when using machine translation

Do not upload sensitive data.

 

Work with Google Cloud Translate -> UI
DeepL

Get help with machine translation!


Post-edit

Make sure that

  • Translation is correct
  • Tags are correct
  • Terminology is applied
  • Style is appropriate

Here you need the translation memory for editing.


Retrofit translation files

Get the translation back where they belong

import texts into files

convert if necessary


Review

Make sure that the final product is correct and looks good.


Cleanup

  • Save translation memories
  • Update terminology files
  • Documentation: Take note of best practices for future usage

Other services

  • Do you want to do these tasks by yourself?

Training can be provided to teach you how.

  • Do you want to know more about privacy?

Privacy (Encryption, Veracrypt, Cleaning, Cloud vs. local)

  • Do you want to know more about data protection?

Data protection (Backups, Minimize)

  • Do you want to know more about getting back you lost data?

Data recovery