Back to Reference
Contents
Reference
511 words
Jun 15, 2026

AtoM Heratio — IIIF AI Extraction

Feature Overview

Component: IiifAiService (ahgAIPlugin) Version: 2.8.2 Category: AI & Automation / Digital Preservation Publisher: The Archive and Heritage Group (Pty) Ltd

Summary

The IIIF AI Extraction module bridges the IIIF Viewer plugin (ahgIiifPlugin) with the AI Services plugin (ahgAIPlugin), enabling automated metadata extraction from digitized archival images. It provides a unified pipeline that orchestrates OCR, Named Entity Recognition (NER), translation, summarization, and face detection against IIIF-served images via an external AI server.

Key Features

Automated Extraction Pipeline

  • OCR — Extracts text from digitized documents (manuscripts, typed records, printed material) using Tesseract via the external AI server
  • Named Entity Recognition — Identifies persons, organizations, places, and dates from OCR text using spaCy NLP models
  • Translation — Translates extracted text between language pairs using Argos Translate
  • Summarization — Generates concise summaries of extracted text content
  • Face Detection — Detects human faces in photographic material for potential authority record matching

Architecture

  • External AI Server: All AI processing is performed by a dedicated server (http://<host>:5004/ai/v1), keeping AtoM's web server lightweight
  • Bridge Pattern: IiifAiService composes existing services (NerService, IIIF annotation services) rather than duplicating functionality
  • Configurable Pipeline: Each extraction step can be enabled/disabled independently per request
  • Batch Processing: Supports processing all digital objects for an information object in a single call

Data Storage

  • OCR text stored in iiif_ocr_text and iiif_ocr_block tables (existing IIIF infrastructure)
  • NER entities stored in ahg_ner_entity table (existing AI infrastructure)
  • Annotations stored in iiif_annotation and iiif_annotation_body tables (W3C Web Annotation model)
  • Extraction history tracked in ai_iiif_extraction table (new)

Settings Integration

  • AI server URL, API key, and timeout configurable via Admin > AHG Settings > AI Services
  • Settings consolidated into ahg_ai_settings table (single source of truth)
  • Backward-compatible fallback to legacy ahg_ner_settings table during migration period
  • Per-feature toggles for OCR, NER, translation, summarization, and face detection

Extraction Tracking

  • Every extraction operation logged with status, duration, error messages, and result counts
  • Status dashboard shows extraction coverage across the collection
  • Statistics API for reporting on AI processing activity

Database Tables

Table Purpose
ai_iiif_extraction Extraction job log — tracks each pipeline run per digital object
ahg_ai_settings Consolidated AI settings (feature + setting_key + value)
iiif_ocr_text Full OCR text per digital object (existing)
iiif_ocr_block OCR text blocks with coordinates (existing)
ahg_ner_entity Extracted named entities (existing)
iiif_annotation W3C Web Annotations (existing)

Technical Requirements

Requirement Details
AtoM Heratio v2.8.2+ with atom-framework
Plugins Required ahgAIPlugin, ahgIiifPlugin
External AI Server Python server with spaCy, Argos Translate, Tesseract
AI Server Endpoint http://<host>:5004/ai/v1
PHP 8.3+
MySQL 8.0+

Standards Compliance

  • W3C Web Annotation Data Model — NER and face detection results stored as IIIF annotations
  • IIIF Presentation API 3.0 — Compatible with IIIF manifests and canvases
  • OAIS — Extracted metadata enriches Archival Information Packages (AIPs)

Use Cases

  1. Bulk OCR of digitized collections — Process thousands of scanned documents to make them full-text searchable
  2. Automated entity extraction — Identify people, places, and organizations mentioned in archival documents
  3. Multi-language collections — Translate extracted text for cross-language discovery
  4. Photographic archives — Detect faces in historical photographs for authority record linking
  5. Collection summarization — Generate concise descriptions of lengthy archival documents

AtoM Heratio is developed by The Archive and Heritage Group (Pty) Ltd for GLAM and DAM institutions worldwide.