AI for National Libraries

AI for National Libraries is SEENL's hub for practical guidance, tools, and training to adopt AI safely, ethically, and effectively across South-East European national libraries. We focus on impact, inclusion, and open standards, so institutions of any size can participate.

Goals & Audience

• Enable responsible AI in cataloguing, discovery, preservation, and user services.
• Share policies, templates, and evaluations to reduce duplication.
• Improve accessibility and multilingual support.
• Build skills through a regional working group and short pathways. For: directors, heads of cataloguing/IT/digital/user services/preservation, policy/legal, and partners.

Priority Areas (2025)

• Metadata enrichment: entity recognition, subject suggestion with human review.
• Multilingual OCR & layout: tuned for regional scripts and historical materials.
• Discovery & Q&A: retrieval-augmented search with citation and librarian triage.
• Accessibility: read-aloud, summarisation, EPUB-A11y/PDF-UA workflows.
• Governance & procurement: DPIA, audit trails, vendor due diligence.

We also surface information-literacy contexts, from media claims to conflict narratives, and regional AI use cases such as AI skin analysis to inform ethics, privacy, and citizen-facing services.

What You Get (Reusable)

• Project charter and DPIA templates.
• Model/prompt evaluation checklist (quality, bias, hallucination).
• Accessibility impact template (screen-reader flows, alt formats).
• Procurement Q&A for security, IP, SLAs, exit.
• Risk register with controls and ownership.
• Guides: RAG/discovery; multilingual OCR.

Working Group & Training

• Cadence: monthly clinic (peer troubleshooting), quarterly roundtable, annual workshop alongside the AGM.
• Pathways: Foundations (ethics & use cases), Implementer (data, eval, oversight), Lead (strategy, procurement, risk).
• Certificates issued upon completion.

90-Day Roadmap

1-30: inventory tasks, select a low-risk pilot, complete charter + DPIA. 31-60: build ground truth, run side-by-side evaluations, set acceptance thresholds. 61-90: limited launch with human-in-the-loop, monitor & publish a short case note.

Metrics & Principles

Indicators: OCR accuracy; entity/subject precision-recall; citation coverage; DPIA completion; accessibility format counts; trained staff; pilots completed. Principles: human oversight; transparency & logs; privacy/security by design; fairness & multilingual inclusion; open standards; sustainability.

Get Involved

Submit a case/tool, volunteer as a reviewer, propose a clinic topic, or partner with us. Contact: SEENL Secretariat / AI Working Group (role, institution, interest area).

FAQ

Will AI replace librarians? No, AI assists; librarians decide. What infra is required? Start with policy + evaluation; pilot on small, scoped datasets. How do we handle many languages/scripts? Evaluate per language/script and prioritise the weakest.

Calls to Action

• Explore the resource library and copy templates.
• Join the working group to shape guidance.
• Propose a pilot and share results.