Skip to content

9.3. Private Document Translation

When translating contracts, internal memos, medical records, or any document containing sensitive information, sending text to a cloud translation service introduces a data privacy risk. Backend.AI GO's built-in Translation feature runs entirely on your machine, so your content never leaves your device.

This guide walks through setting up a private translation workflow—from choosing a model to creating glossaries and translating files—all without an internet connection.

Why Translate Locally?

Concern Cloud translation Backend.AI GO
Data privacy Text is sent to external servers Text stays on your machine
Compliance May violate data-handling policies Fully local, auditable
Internet required Yes No
Custom terminology Limited or paid add-on Built-in glossary support

Prerequisites

Before you begin, make sure you have:

  • Backend.AI GO installed and running
  • A translation-capable model downloaded — see Compatible Models below
  • Enough disk space and RAM for the model (check model requirements on the download page)

Step 1: Download a Translation Model

  1. Open Backend.AI GO and go to the Models page.

  2. Search for a translation model. TranslateGemma 4B Instruct is recommended for its accuracy and language coverage.

  3. Click Download and wait for the model to finish downloading.

  4. Once downloaded, load the model so it's ready for inference.

General-Purpose Models Work Too

Models like Llama or Qwen can also translate, but dedicated translation models typically produce more accurate and natural results, especially for less common language pairs.

Gated Models May Require a Hugging Face API Key

Some models (such as TranslateGemma) are gated models on Hugging Face, meaning you must accept the model's license terms and authenticate to download them. If the download fails or you see an access error, go to Settings → General and enter your Hugging Face API token. You can generate a token at huggingface.co/settings/tokens.

Step 2: Translate Text

Text translation Text translation

  1. Click the Translation icon in the sidebar to open the translation interface.

  2. In the Text tab, select your loaded translation model from the dropdown.

  3. Set the Source Language (or leave as "Auto-detect") and the Target Language.

  4. Paste or type your text in the left panel.

  5. Press Cmd+Enter (macOS) or Ctrl+Enter (Windows/Linux) to translate.

  6. The translated text appears in the right panel. Click the copy icon to copy it to your clipboard.

Real-Time Mode

Real-time translation Real-time translation

For interactive sessions—like drafting bilingual emails—toggle Real-time mode in the top-right corner. Translation happens automatically as you type. Disable it for longer texts to avoid unnecessary processing while you're still composing.

Step 3: Set Up a Glossary

When translating technical, legal, or medical documents, certain terms must always be translated the same way. Glossaries enforce this consistency.

Create a Glossary

Empty glossary tab Empty glossary tab

  1. Go to the Translation page and open the Glossary tab.

  2. Click Create Glossary.

  3. Fill in the details:

    • Name: A descriptive name (e.g., "Legal Terms EN→KO").
    • Description: Optional context about the glossary's purpose.
    • Source Language: The language of the original terms.
    • Target Languages: One or more translation target languages.

    Creating a glossary Creating a glossary

  4. Click Save Changes.

Add Terms

  1. Click Edit Glossary on the glossary card you just created.

  2. Click Add Entry to add a new term row.

  3. Enter the source term and its translation for each target language.

    Adding a new term Adding a new term

  4. Repeat for all terms you want to standardize.

Example glossary for a software company:

Source (EN) Target (KO)
Backend.AI Backend.AI
inference 추론
model 모델
fine-tuning 파인튜닝
endpoint 엔드포인트

Apply a Glossary to Translation

Translation with glossary applied Translation with glossary applied

  1. Go to the Text or File tab.

  2. Select a glossary from the Glossary dropdown next to the model selector.

  3. Translate as usual. The model will use the glossary terms to ensure consistent translation.

Glossary Scope

A glossary applies to both text and file translation. You can switch between glossaries at any time without reloading the model.

Step 4: Translate Files

File translation File translation

For translating entire documents rather than snippets of text:

  1. Open the File tab in the Translation page.

  2. Select your model and configure source/target languages.

  3. Drag and drop a file onto the drop zone, or click to browse.

  4. Supported file formats:

    Format Notes
    TXT, MD Plain text and Markdown files
    DOCX Microsoft Word documents
    PDF PDF documents
    PNG, JPG, WebP Images — text is extracted via OCR before translation

  5. The translated output is displayed below the upload area.

Batch Workflow

For multiple files, translate them one at a time. Copy each result before uploading the next file.

Example Workflow: Translating a Contract

Here is a complete workflow for translating a confidential contract from English to Korean:

  1. Prepare: Download and load TranslateGemma 4B Instruct.

  2. Create a glossary: Add legal terms that must remain consistent (e.g., "indemnification" → "면책", "jurisdiction" → "관할권").

  3. Apply the glossary: Select it from the Glossary dropdown on the Text or File tab.

  4. Translate: Upload the contract PDF in the File tab, or paste sections into the Text tab for review.

  5. Review: Check the translation for accuracy. Adjust glossary terms if needed and re-translate.

All of this happens locally. The contract text never leaves your machine.

Compatible Models

Model Description
TranslateGemma 4B Instruct Dedicated translation model with broad language support. Recommended.
Llama (various sizes) General-purpose model capable of translation.
Qwen (various sizes) General-purpose model with strong multilingual capabilities.

Dedicated translation models generally produce better results than general-purpose models, particularly for:

  • Less common language pairs
  • Technical or domain-specific content
  • Maintaining natural phrasing and grammar

Tips

  • Disconnect from the internet if your security policy requires it. Backend.AI GO's translation works fully offline once the model is loaded.
  • Create separate glossaries for each project or domain (legal, medical, technical) to keep terminology organized.
  • Use Real-time mode for short phrases and manual mode for longer documents to control processing load.
  • Review translations before using them in official documents. AI translation is a powerful aid, but human review ensures quality.