Templates

Owned by Konstantin Dranch
Last updated: Apr 12, 2024
6 min read

Follow the steps described in the article Quickstart, then proceed to the next steps.

Creating Templates

The new templates add an ability to chain in a GPT prompt before and after machine translation. The text of the segment in the prompt is indicated with a variable {text}.

Please note that OpenAI's model won't "see" the context, only the current segment text. For example, when post-editing MT output, OpenAI won't yet be able to see the source, these functions are being built for later releases. 

In order to try new functionality, you need to go to the Translation - Templates section of the portal.

Press New Template and name the template you are creating, then press Create. You can give any name to the template but try to keep the template name meaningful so that you won’t get confused later.

Point your mouse to the plus sign on the left side of the page , you will see the options available for adding to your prompt.

Pre-editing

Pre-editing option gives you the ability to change the source text before sending it to translation.

You can see below a few examples of Pre-editing prompts:

  1. Rewrite in the Simplified Technical English {text} 

  2. Rewrite for clarity

  3. Change units of measure from Imperial to Metric system in the text here:  {text} 

  4. Proofread the text {text} 

  5. Rewrite the following in inclusive language: {text} 

Press the Add button to add the Pre-editing step.

In the Pre-editing Setup window, you can select one of the prompts from the preset list or one of the prompts you created in Prompt Studio.

If needed, you can remove Pre-editing, Translation, or Post-editing steps by selecting the step using your mouse and clicking the Bin sign .

 

Translation

If you want to create a template for translation, use the +ADD sign on the right of Translation.

You will need to select the Source and Target language, then select the First engine from the drop-down list. It is better to choose the option All Domains as it will allow the MT engines to provide better-suited translation.

There is an option to Add a secondary engine. You can use it when you need to make sure that translation will be provided uninterrupted, so even if the first MT Engine is experiencing any technical difficulties, you will still be able to receive translation via the second MT engine.

Press Save to save your template.

If you would like to delete the Translation step, select the step using your mouse and click the Bin sign .

 

Post-editing

If you would like to make any changes to the translated text, you can add the Post-editing step to your template.

Ideas for prompts - after machine translation

  1. Terminology superimposition. "Rewrite the text using the term X instead of Y, do no other changes: {text}" 
    Nota bene: larger glossary support is being developed

  2. Tone of voice: friendly, formal, etc

  3. Formality (toi, vous)

  4. Rewrite imitating the style of Hemingway: {text}

  5. Correct errors in machine translation {text} 

  6. Rewrite the translation so that the length does not exceed 55 characters.

To add a Post-editing step to your template, use the +ADD sign on the right of Post-editing.

Select the prompt from the Prompt drop-down list and then press Save.

 

 

You are all set! The template is ready to be used.

Tag Handling

Depending on the type of text you will be translating, you can add a Tag Handling step. Press +ADD on the right of the Tag Handling option.

You can select any filters from the available options depending on the type of text you are translating. By default, no filters will be applied.

  • Remove Tags: To be used to remove tags in target text while they are present in source text.

  • HTML tags keeper: To be used by specific clients only. It is advised to use HTML Keeper ver 2.0 in other cases. This filter should be used when the source files are HTML files containing the tags and programming instructions.

  • Tag spaces cleaner: Removes spaces before and after the tags in the target text.

  • Unsplit Google translation with tags: Trados only. To be used when inline tags are used in the text. In such cases without this filter, separate words between tags are translated as stand-alone words, translation will not be fluent. Can not be used together with HTML tags keeper ver1/ver 2.0.

  • Space Keeper: Checks if the spaces between and after the tags are present in translation, and if they are not, it adds the spaces.

  • Tags encoder: Tags are handled in a way similar to the one in RWS Trados, where tags are encoded to tg1, tg2, etc. Provides more fluent translation by converting the tags into one short tag, thus text can be translated correctly. It is not recommended for HTML files, use HTML Keeper ver 2.0 instead.

  • Units of measure corrector: To be used when English is a source language. Used to keep units of measure intact after translation, also adds a space between the number of units and the unit abbreviation. For example, 5GB will be translated as 5 GB. Used per client’s request, can be applied if technical specifications are translated.

  • Alternative dialect: For the GPT engine, when any language dialect is required, for example, convert Spanish-Spain into Spanish-Latin American dialect.

  • HTML Keeper ver 2.0: This filter should be used when the source files are HTML files and they contain HTML tags. This filter parses the source text and extracts the text, then sends the text for translation. After getting the translation back, the tags are inserted back. The logic behind this filter helps in getting a more fluent translation.

  • DNT tags: This option will allow you to set words which translator will ignore. For this, use the translate=“no” attribute.