AI Translator Functional Boundaries

This document describes the functional boundaries, known formatting issues, and limitations for translating various types of documents.

1. PDF Documents (Non-scanned)

Unsupported document types

• Design documents (drawings, posters)

• Right-to-left layout documents (such as Arabic)

• Vertical text

• Documents using Type3 fonts

Possible issues

Download format description

Default option:

• Only supports downloading in PDF format

Optional configurations (must be enabled by an administrator):

1. Built-in converter (free):

   • Can be converted to Word (.docx) format

   • ⚠️ Text loss or blank pages may occur

   • Suitable for scenarios requiring simple editing

2. Adobe service (paid):

   • High conversion quality with good format retention

   • Requires separate service purchase

   • Suitable for scenarios with strict formatting requirements

Special notes:

• Scanned documents do not support docx conversion

• PDFs converted from PPT and then converted to docx may have formatting issues

Scanned PDF

Applicable scenarios

• ✅ Scanned paper documents

• ✅ Well-formatted document photos

• ✅ Clear scans of contracts and reports

Unsuitable scenarios

• ❌ Scans with watermarks

• ❌ Blurry photos

• ❌ Handwritten documents (low recognition rate)

• ❌ Partial vertical text within pages

Possible issues

Special features

• Automatic orientation detection: Supports 0°/90°/180°/270° rotation with automatic correction

• Bilingual output: Can generate both translation-only and bilingual comparison versions

• Pagination processing: Documents over 50 pages are automatically processed in batches to ensure stability

Technical requirements

• OCR service: Azure OCR is recommended (more accurate recognition)

• If not configured, RapidOCR will be used automatically (lower quality)

3. Word Document Translation

Applicable scenarios

• ✅ Contracts, agreements

• ✅ Proposals, reports, papers

• ✅ Operation manuals, training materials

Possible issues

Feature description

Supported for translation

• ✅ Main body content

• ✅ Heading styles

• ✅ Table text

• ✅ SmartArt text content (styles may be simplified)

Content not translated

• ❌ OLE embedded objects (embedded Excel, Visio, etc.)

• ❌ Revision marks and comments

• ❌ Field codes (dates, cross-references, etc.)

Notes

• Online preview is not currently supported after translation; download to view

• Complex SmartArt styles (gradients, special effects, etc.) may be simplified

4. Excel Document Translation

Applicable scenarios

• ✅ Data tables, lists

• ✅ Statistical reports

• ✅ Process description tables

File opening prompt

A "repair needed" prompt may appear when opening for the first time:

• ✅ This is normal

• ✅ Click "Yes" or "Repair"

• ✅ Content will not be lost

Reason: Complex styles or charts may trigger Excel format validation when written back

Intelligent skip rules

The system automatically skips content that does not need translation:

• 🔢 Formulas: Cells starting with = remain unchanged

• 📅 Dates: Date-type cells are not translated

• 🔢 Pure numbers: To avoid damaging data

• 📏 Merged cells: Only the top-left content is translated

Functional limitations

Worksheet name rules

• Maximum 31 characters (Excel system limitation)

• Excess parts are automatically truncated

• Cannot contain: / \ ? * [ ] :

Notes

• Formula calculation results will only display after reopening the file

• Charts in files may be lost after translation

• Complex conditional formatting may be simplified

5. PowerPoint Document Translation

Applicable scenarios

• ✅ Presentations

• ✅ Training courseware

• ✅ Product introductions

Possible issues

Feature description

Supported for translation

• ✅ Slide body text

• ✅ Titles and subtitles

• ✅ Text box content

• ✅ Table text

• ✅ SmartArt text (with standard structure)

• ✅ Chart titles, axis titles, legends

• ✅ Notes page content

Content not translated

• ❌ Section names (only content visible during slideshow is translated)

• ❌ Comments

• ❌ Embedded video/audio

• ❌ Complex third-party embedded charts

Content not affected

• ✅ Animation effects

• ✅ Transition effects

• ✅ Hyperlinks

Font compatibility option (recommended)

Background:

• PPT font compatibility is relatively weak

• Garbled text may appear after translation

Feature description:

• When the target language is Chinese, Japanese, or other East Asian languages

• Automatically replaces fonts in regular text boxes and tables

• Chinese is automatically replaced with "Microsoft YaHei"

Whether to enable:

• ✅ Recommended to enable: Automatic handling, saves time and effort

• ❌ Do not enable: Fonts at garbled positions need to be replaced manually

Notes

• Online preview is not currently supported after translation; download to view

• If the translation is longer than the original, it is recommended to check text box sizes

6. General Limitation Notes

Common characteristics across all formats

1. Translation length changes

• Length differences across languages:

  • Chinese → English: usually 20-50% longer

  • English → Chinese: usually 30-40% shorter

  • Chinese → German: usually 40-60% longer

• Possible resulting issues:

  • Text box overflow

  • Pagination changes

  • Incomplete display of text in table cells

2. Automatic skip rules

• ✅ Pure numeric content (e.g.: 12345)

• ✅ Pure symbol content (e.g.: ***、---)

• ✅ Formulas and dates

• ✅ Content with OCR confidence below 50% (scanned documents)

3. Terminology consistency

• Supports custom glossaries

• Translation quality is related to model capability

• Mismatches may occur when model capability is weaker

4. Format retention mechanism

• By default, formatting at the beginning position of paragraphs is retained

• Formatting is not preserved precisely for every character

• Possible outcome: some formatting may be lost or may not follow the translated text

Example:

Original: This is a paragraph containing 【bold】 and *italic* text
Translation: It may all be displayed in plain formatting

5. Long text processing

• Extra-long single paragraphs are automatically split for translation

• In rare cases, transitions across segments may be less natural

6. Translation timeout notes

Reference data (using GPT-4 as an example):

Timeout handling:

• Documents over 500,000 characters are recommended to be split

• Translate in batches and then merge

7. Best Practice Recommendations

Before translation

1. Document preparation

• ✅ Clean up unnecessary formatting and styles

• ✅ Delete pages that do not need translation

• ✅ Check whether the document is corrupted

• ✅ Ensure scanned documents are clear

2. Choose the appropriate format

• Pure text content: Prefer Word

• Contains a large amount of data: Use Excel

• Presentation materials: Use PPT

• Paper documents: Scan and use scanned document mode

3. Handling very large files

• Documents over 500,000 characters are recommended to be split

• Divide into multiple files by chapter or topic

• Merge after translation

During translation

1. Option configuration

• PPT translation: It is recommended to enable "Font Compatibility"

• Scanned documents: Choose "Bilingual Output" for easier verification

• Terminology-sensitive content: Configure a glossary

2. Be patient

• Large file translation takes time

• Avoid repeated submissions

• You can switch to background execution

After translation

1. Required checks

• ✅ Open and review with the corresponding software (Excel/Word/PowerPoint)

• ✅ Click "Yes" if prompted with "repair needed"

• ✅ Word table of contents needs to be updated (right-click → Update Field)

• ✅ Check whether key terminology is accurate

2. Formatting adjustments

• ✅ Check whether text boxes overflow

• ✅ Adjust pagination positions

• ✅ PPT font size may need to be reduced

• ✅ Excel column widths may need adjustment

3. Content verification

• ✅ Compare important paragraphs against the original

• ✅ Confirm that numbers and dates have not been changed

• ✅ Ensure formula calculation results are correct (Excel requires reopening)