Translator Notes

Providing additional context to improve translations

Introduction

Translation is not simply a matter of converting words into other words. The context and intent of words affects how they should be translated.

For example, the word "light" can refer to illumination, the weight of a font, or a color theme. These different meanings must be considered for accurate translation.

To provide this kind of context, some types of buckets support translator notes. These are short descriptions that describe things like:

  • What the content is meant to communicate
  • Where and how the content appears in your application
  • Product-specific terminology or tone requirements

This context can then help AI models translate content more effectively.

How It Works

Certain types of buckets, such as JSONC, support comments. When content is sent to an LLM for translation, these comments are included in the request and can be used to affect how the content is translated.

For example, in the following code snippet, a comment eliminates the ambiguity of the word "Records":

{
  // Medical context: refers to patient medical records, not music/audio
  "records": "Records",
}

Supported Buckets

Translator notes are only supported in certain types of buckets.

The supported buckets include:

The exact syntax for providing translation context depends on the bucket. To learn more, see the linked documentation.