User Guide

The DNNTC Viewer is a transcription and verification interface for working with digitized tables and character-level OCR data. This guide explains how to navigate volumes, inspect scanned images, search text, and submit corrections.


1. System Overview

The interface is divided into three synchronized panels:

  • Left Panel: Original scanned page image (IIIF viewer)
  • Middle Panel: Structured transcription grid
  • Right Panel: Search and navigation tools

All three panels are linked by shared volume/page state, meaning navigation updates all views simultaneously.

Screenshot 1: Full interface overview showing all three panels

2. Navigation

Volume Selection

Volumes represent separate manuscript or dataset collections. Use the dropdown menu in the middle panel to switch between volumes.

  • “Prev” and “Next” buttons cycle through volumes sequentially
  • Switching volume resets the page index automatically
Screenshot 2: Volume dropdown and navigation buttons

Page Selection

Each volume contains multiple pages. You can navigate using:

  • Page dropdown selector
  • Previous / Next page buttons

Page changes synchronize:

  • Left image viewer updates (scanned page)
  • Middle transcription grid updates

3. Image Viewer (Left Panel)

The left panel displays the original scanned manuscript page using a deep zoom viewer. You can pan and zoom freely using mouse or trackpad gestures.

  • Scroll to zoom
  • Click + drag to pan
  • Double click to zoom in
Screenshot 3: Image viewer zoomed into a character region

4. Transcription Grid (Middle Panel)

The transcription grid displays structured OCR output as a spatial character layout. Each cell corresponds to a row and column in the original table structure. Characters are individually selectable, which allows you to create a suggestion report.


5. Stack Mode (Document View)

Stack Mode changes the interface from single-page view to continuous scrolling document mode. Instead of switching pages, all pages are rendered vertically in sequence.

When to use Stack Mode

  • Reading across multiple pages
  • Contextual transcription review
  • Searching across full volume layout

Behavior

  • Pages load progressively in vertical blocks
  • Scrolling automatically updates the active page
  • Image viewer syncs with visible page section
Screenshot 5: Stack Mode active with multiple pages visible

6. Search System

The search panel allows full-text searching across all volumes. Results are matched against the underlying character corpus.

Search Workflow

  1. Enter a query in the search box
  2. Press Enter or click Search
  3. Review matched snippets
  4. Click a result to jump to the corresponding page

Search Results

  • Highlighted snippets show surrounding context
  • Results are grouped by volume, page, and table
Screenshot 6: Search results panel with highlighted match
Show snippet highlighting and a clickable result entry.

7. Character Correction System

The correction system allows users to report transcription errors at the character level. This is a key part of dataset refinement and validation.

How to Submit a Correction

  1. Click a character in the transcription grid
  2. Click “Suggest Correction”
  3. Enter corrected character and optional notes
  4. Submit the correction

Correction Data Captured

  • Original character
  • Suggested correction
  • Reason category
  • Row/column metadata
  • Source file and page context
Screenshot 7 and 8: Correction modal open

8. Downloading Corrections

All submitted corrections are stored temporarily in the session and can be exported.

  • Click “Download” to export JSON file
  • File includes timestamped correction log
Screenshot 9: Download button active with pending corrections

9. Tips & Best Practices

  • Use Stack Mode for long-range comparison across pages
  • Use single-page mode for precise character correction
  • Always verify against the original image before submitting corrections
  • Use search to locate recurring patterns or terms

10. System Notes

This interface is designed for high-resolution transcription validation workflows. All edits are non-destructive and stored as correction metadata only.