The Engine: IIIF Paleography

The core of this exhibit is powered by IIIF Paleography, an open-source utility developed to make displaying the transcription and the reasoning easy.

While IIIF (International Image Interoperability Framework) provides a powerful way to deliver high-resolution images, it does not inherently "know" what is written on them. This tool acts as the translator.

What is IIIF Paleography?

IIIF Paleography is a Python-based utility designed to automate the transcription and analysis of historical documents. It interacts with the IIIF Presentation API to fetch document manifests, extract their individual canvases, and process them through the Google Gemini API.

The Technical Workflow

  1. Manifest Ingestion: The utility accepts a IIIF Manifest URL (e.g., a letter from a Texas A&M President).
  2. Canvas Processing: It iterates through every page (canvas) in the manifest.
  3. AI Orchestration: It sends the high-resolution image link to Google Gemini along with a highly tuned System Prompt.
  4. Annotation Generation: It receives two types of data back from the AI:
    • Transcript: The literal text.
    • Commentary: The reasoning behind the transcription.
  5. Standardized Output: The utility packages these results into W3C Web Annotations, making them compatible with modern IIIF viewers like Clover, Theseus Mirador, or Universal Viewer.

Key Features

Standards-Compliant Output

By generating valid IIIF Annotation Pages, the utility ensures that the transcriptions aren't just stuck in a text file. They are tethered to the specific image coordinates, allowing for features like:

  • Searchable text overlays.
  • Side-by-side view of original ink and digital text.
  • Accessibility support for screen readers.

Semantic Markup

Unlike basic OCR (Optical Character Recognition), this utility instructs the AI to use semantic HTML tags allowed by the IIIF spec (such as <p>, <span>, and <b>). This preserves the structural integrity of the letter—distinguishing between the body of the text, signatures, and marginalia.

Automated Reasoning

The utility specifically requests the "Chain of Thought" reasoning that powers our Commentary section. This feature is what transforms the tool from a simple transcriber into a digital paleographer.

Integration in this Exhibit

In this exhibit, we have used this utility to process the collections of Texas A&M's early presidents. Every transcription you see was generated using this pipeline, demonstrating a scalable way to bring "dark archives" into the light of the digital age.