📄 Metadata Upload Guide

You can improve the accuracy and relevance of automated processing by uploading a small JSON file with additional metadata about your documents. The fields below are optional — include only what you know.

Tip: If you’re unsure about a field, you can leave it out. The system will use defaults automatically.

📝 writing_style

Describes the visual form of the text in the document.

• Options:

  • "handwritten"
  • "printed"
  • "typed"

This helps the system choose between transcription strategies (e.g., handwriting recognition vs. printed text).

🌍 language

The main language(s) used in the document.

• Options:

  • "english"
  • "spanish"
  • "portuguese"
  • "french"
  • "german"
  • "mixed"

📅 time_period

Gives a rough estimate of when the document was created. This can influence how the system interprets spelling, abbreviations, and writing style.

• Options:

  • "contemporary"
  • "mid_20th_century"
  • "early_20th_century"
  • "19th_century_or_earlier"

⚙️ transcription_preferences

Specify how you’d like the system to handle certain aspects of transcription. Each setting is optional — include only the ones you want to customize.

• expand_abbreviations: true to expand abbreviations like “Dr.” to “Doctor”
• preserve_line_breaks: true to keep the original line formatting
• retain_punctuation_and_spelling: true to preserve original punctuation and spelling
• normalize_to_modern_language: true to modernize spelling and phrasing
• ignore_marginalia: true to skip notes in the margins

Defaults:

{
  "expand_abbreviations": false,
  "preserve_line_breaks": true,
  "retain_punctuation_and_spelling": true,
  "normalize_to_modern_language": false,
  "ignore_marginalia": false
}

📐 layout_structure

Describes how the text is arranged on the page. This helps guide interpretation of complex layouts.

• Options:

  • "free_form"
  • "paragraphs"
  • "lists"
  • "tables"
  • "forms"
  • "mixed"

🎨 non_textual_elements

List any non-text content that appears in the document. This information can help avoid misinterpreting these as text.

• Options (use an array of values):

  • "illustrations"
  • "stamps_or_seals"
  • "handwritten_notes"
  • "diagrams"
  • "charts_or_graphs"

Example:

"non_textual_elements": ["illustrations", "handwritten_notes"]

🌈 color_format

Indicates the color style of the scanned image. This helps the system calibrate processing.

• Options:

  • "color"
  • "grayscale"
  • "black_and_white"
  • "mixed"

↔️ orientation

Specifies how the page is oriented. This helps avoid mistakes when the image isn’t upright.

• Options:

  • "portrait"
  • "landscape"
  • "square"

🧾 image_quality_notes

List any known image quality issues so the system can adapt or flag the content.

• Options (use an array of values):

  • "blurry"
  • "low_contrast"
  • "partial_page"
  • "includes_fingers"
  • "includes_color_checker"
  • "damaged_or_stained"

Example:

"image_quality_notes": ["blurry", "includes_fingers"]

✍️ description

A free-text description of the document set. This can provide general context to improve results for both transcription and metadata generation.

Example:

"description": "These are scanned forms from a late 20th-century survey project on urban housing."