✂️ Text segmentation

When you upload a document to Redokun, the system automatically processes the text to optimize it for translation. This involves segmenting the text into manageable units and applying filters to exclude non-translatable elements.


💡 What Is Text Segmentation?

Text segmentation is the process of dividing written text into meaningful units, such as words, phrases, sentences, paragraphs, or topics.

When you upload a document to Redokun, the software runs a process called segmentation that breaks the text into smaller "segments".

The goal of creating these segments is to make translation easier and faster.

Here's how:

  • The translator can focus on translating smaller units of text rather than having to look at the entire written content.
  • Having smaller logical units of text makes it easier to reuse translations in future projects.

📄 Segment Granularity Options

Redokun offers two segmentation options:

  • Paragraph Segmentation: Divides text into paragraphs.
  • Sentence Segmentation (default): Further divides paragraphs into individual sentences.

🧭 Choosing the Right Segmentation

Your choice of segmentation method can depend on whether you're prioritizing AI-based translation or Translation Memory (TM) reuse.

  • Use Paragraph Segmentation if you're relying on AI-generated translations. Full paragraphs give AI models the context they need to generate more fluent, accurate, and natural-sounding output.
  • Use Sentence Segmentation if you're building or leveraging a Translation Memory. Sentence-level units maximize reuse across projects and improve consistency in terminology and phrasing.
  • Paragraph Segmentation is ideal for non-professional translators, as it presents larger, more digestible text blocks.
  • Sentence Segmentation is recommended for professional translators or when working with translation vendors, as it enhances the reusability of translations and improves consistency.

✨ AI Translation and Paragraph Segmentation

AI translation tools (like DeepL, Google Translate or ChatGPT by OpenAI) tend to produce better results when working with larger chunks of text. That’s why paragraph segmentation can improve translation quality in some cases:

  • More context: Full paragraphs give the AI a better sense of meaning, improving word choice and tone.
  • Natural flow: Paragraphs help AI preserve sentence structure and readability.
  • Better coherence: When style and voice matter (e.g., in marketing copy), working with paragraphs helps maintain consistency.

Use Paragraph Segmentation when tone, nuance, and fluency are critical — and your team plans to rely heavily on AI.

📚 To improve AI translation quality further, consider creating a Glossary. Learn more about the Glossary feature →


⚙️ How to Change Segmentation Settings

To select your preferred segmentation type:

  1. Click on your name in the upper-right corner and select Settings.
  2. Navigate to the Segmentation tab.
  3. Choose either Sentence or Paragraph segmentation.
  4. Click Save to apply your changes.

⚠️ Changes to segmentation settings will only affect documents uploaded after the change.


📘 Best Practices for InDesign and Word Documents

For optimal results when working with InDesign or Word files:

  • Ensure consistent use of styles and formatting to facilitate accurate segmentation.
  • Avoid unnecessary manual line breaks; use soft returns where appropriate.
  • Refer to our comprehensive resources for detailed guidance:

↩️ Soft Returns and Segment Wrapping

Redokun creates a new segment whenever it detects a hard return (Enter) or a tabulation.

To control segment breaks within a paragraph:

  • Use soft returns (Shift + Enter) to insert line breaks without creating new segments.
  • To make soft returns visible to translators, enable the Convert Soft-returns option:
    1. Go to Settings > Advanced Settings.
    2. Check the box for Convert Soft-returns.

ℹ️ For more information on soft returns, refer to Wikipedia's explanation.

Did this answer your question? Thanks for the feedback There was a problem submitting your feedback. Please try again later.

Still need help? Contact Us Contact Us