AUMLayer · Institutional manuscript guide
A practical manuscript digitisation evaluation checklist
Move from a broad digitisation ambition to requirements that an institution can discuss, test and accept.
A useful manuscript brief describes what people should be able to do with the collection. It then connects those outcomes to source material, review responsibilities, digital deliverables and a workable handover. That makes the discussion clearer for custodians, scholars and technology teams alike.
This AUMLayer checklist helps structure that discussion before a LipiForge programme is scoped. It is technical planning guidance: the institution retains ownership of its procurement process, applicable terms and approval decisions. The matrix below can be used to compare the same requirements across proposed workflows.
What should a manuscript digitisation brief specify?
Specify the source inventory, languages and scripts, desired text and edition outputs, scholarly review roles, source-linking requirements and operational handover. For each requirement, identify why it matters, what evidence to inspect and who accepts it. A LipiForge discussion can then connect the collection’s needs with recognition, review and delivery scope.
Use a requirement-and-evidence matrix
The matrix is a starting point for a collection-specific brief. Replace the suggested acceptance roles with the people authorised by your institution. Record unresolved choices explicitly so that a sample demonstration and a later delivery are assessed against the same expectations.
Keep image capture, recognition, scholarly correction, writer-style reconstruction and physical printing distinguishable. A programme can combine them, but each output has a different purpose and requires appropriate evidence.
| Requirement | Why it matters | Evidence to request | Suggested acceptance owner |
|---|---|---|---|
| Source inventory | Avoid missing or duplicated material | File/page reconciliation and original identifiers | Collection custodian |
| Languages and scripts | Match the work to actual material | Representative hands, layouts and difficult pages | Subject specialist |
| Recognition output | Assess the proposed reading | Region-linked transcription and error examples | Scholarly review lead |
| Correction workflow | Keep decisions traceable | Source reading, proposal and approval history | Editorial authority |
| Search and metadata | Retain context in discovery | Queries leading to the correct page and text version | Library or research lead |
| Edition preparation | Serve the intended audience | Reading/visual proofs in agreed formats | Publication owner |
| Private deployment | Define operational responsibility | Access, dependencies, retention and recovery plan | Institutional IT owner |
| Handover | Keep outputs useful after delivery | Sample export opened in the receiving environment | Joint collection and IT owners |
Define the sample before judging the result
A representative sample should reflect the work to be delivered: different writers, layouts and image conditions, alongside ordinary readable pages. Agree how the reference reading is prepared and how uncertainty is recorded. Keep the acceptance material distinct from material used to improve or demonstrate the workflow.
Report recognition and editorial review separately. A corrected passage may be valuable, but the institution should also understand the effort and decisions that produced it. For writer-style editions, assess textual correctness and visual relationship to the source as separate questions.
Make milestones meaningful to the collection owner
A page count is useful only when its state is clear. Received files, recognised pages, reviewed readings and released editions describe different progress. Agree how a batch is accepted and how rejected or incomplete material returns to the workflow.
Milestones can then be tied to inspectable outcomes, such as an identified group of works with reviewed text and source-linked delivery. This gives the institution a clearer view of progress than a single conversion total and helps define responsibilities when requirements change.
Include the receiving environment in the evaluation
A sample handover should be opened in the institution’s own environment. Check whether text is usable, identifiers remain intact, links open the right pages and the relevant editions can be identified. Include the people who will administer and use the material after delivery.
For private deployment, discuss hosting, access, retention, external dependencies and operational continuity. These requirements are part of the programme rather than an appendix to a text-conversion task. They should be reflected in the actual architecture and accepted responsibilities.
Bring the brief into a LipiForge discussion
AUMLayer can use the collection inventory, representative material and intended outputs to discuss a LipiForge programme. Recognition, scholarly review, source-linked discovery and writer-style editions can then be considered as connected parts of the institution’s work.
The checklist remains a practical tool for defining that work, rather than a claim that a specific procurement requirement has already been met. Formal eligibility and tender compliance are assessed against the institution’s actual documents and required evidence.
Questions institutions ask
- Should the brief name a recognition model before evaluation?
- Start with the required outcomes and collection evidence. A specific technology requirement should have a documented reason and still be evaluated on the intended material.
- Who should approve manuscript text and visual editions?
- The institution should name its editorial and publication authorities. Textual readings, corrections and visual proofs can require different expertise, so their acceptance responsibilities should be explicit.
- Can this matrix replace the institution’s formal tender documents?
- No. It is technical planning guidance that can inform the brief. Procurement terms, eligibility and formal compliance remain governed by the actual process and required evidence.