DH/AI African Studies workshop archive: reference guidance for supervised archival processing
Workshop archive and collective-paper REFERENCE-GUIDANCE L2 for human-directed archival discovery/cataloguing: related OCR/HTR, summary and metadata pathways, expert verification and collection-specific purpose/rights/governance—not one prescribed serial architecture. It is not an archive-executed or empirically evaluated workflow, and cited project outcomes are not inherited.
L2Research workflow (workflow level, not a quality rating) Source: Repository-documented Outcome: Not empirically evaluated here
Research applicability
A critical reference for choosing proportionate AI assistance in archival discovery/cataloguing, considering related text/summary/metadata uses and their expert-verification needs, collection-specific rights/governance and unresolved ontology choices. It supports situated judgement, not one universal architecture, a turnkey implementation or validated effectiveness.
Getting started
Untested editorial guidance: read the original archive README and saved collective-paper passages on intentional practice/no single answer, situated purpose/rights, distinct archival examples, ontology disagreement/participation limits and no universal model. Consider only pathways appropriate to a collection; do not turn the examples into an assumed serial implementation plan. No source software/private-note extraction, publisher/DOI resolution or new candidate retrieval is authorised.
Workflow & human oversight
Reference-guidance L2 only for human-directed archival discovery/cataloguing through related, collection-specific OCR/HTR, summary/triage and metadata pathways. These are not one prescribed serial architecture or archive-executed workflow. Expert verification and purpose/rights/governance are normative; no universal ontology model, implemented safeguards or inherited project outcomes are claimed.
Workflow level is not an evidence score; higher is not better evidence.
Approach
The original archive owns a collective paper that brings together related archival discovery/cataloguing pathways, NOT a universal serial chain. Hofheinz illustrates HTR/content analysis/summary for triage; another project, IWAC, illustrates LLM-assisted OCR and entity/metadata uses. These are distinct project accounts, not consecutive stages of one implemented archive system. Metadata suggestions call for domain-expert verification and leave substantive interpretation to human study. The paper explicitly reports NO CONSENSUS on balancing human-designed ontologies with AI-generated metadata. Collaborative, revisable ontology design is discussed as one approach, not a universal input/prerequisite or achieved participatory-control guarantee; the source also documents skills/funding/power asymmetries. Collection-specific purpose, rights and governance condition choices without prescribing one executable architecture. The deterministic concept graph is separate, and degree is not consensus.
Limitations
Paper-excerpt/section review of saved text, not a complete critical manuscript/PDF or publisher/DOI audit; repository metadata says corrected PDF unavailable. The guidance spans distinct projects and related pathways, not one prescribed sequential pipeline. The paper records NO CONSENSUS on balancing human-designed ontologies and AI-generated metadata; collaborative ontology design is not established as universal or achieved, and a participatory-design example reports skills/funding/power asymmetries. No archive-owned model/API sequence, execution trace, runtime approval/stop/budget enforcement, accuracy or actual outcome evaluation is established. Collection-specific rights/verification recommendations are not measured safeguards. No cited project benefits/results are inherited; private source notes were not accessed, and corrected public graph provenance is not a fresh private-note extraction.
Reproducibility
Source/metadata/excerpt inspection only; no archive build, graph execution, private-note access, OCR/model/API run, cited-project reproduction, DOI resolution or independent scientific outcome evaluation. Own editorial guard/provenance tests do not evaluate the archive's AI workflow or its safeguards.
Adoption requirements
No installation is required to use the saved material as a reference. Any later application requires collection-specific choices about aims, tools, expertise and rights/governance, not proof of one universal chain or established collaborative categories. The paper recommends situated judgement, notes unresolved ontology balance, and supplies counterevidence to assuming equitable participation is achieved. Distinct documented code/site/paper/photograph terms remain separate. Editorial application advice is not a verified implementation feature.
Comments
No comments yet.
Add a comment
Comments are held for moderation before appearing. Critical or corrective comments are welcome — especially evidence that an entry overstates what has been verified.