Plan a pilot box
Menu

Metadata mapping worksheet for Dublin Core

A field-by-field map from your own catalog to the 15 Dublin Core elements, agreed before records are exported, imported or published.

Worksheet

The rows are a fictional example. Edit any cell, add or remove rows, then download a CSV for your team. Nothing leaves your browser.

Your fieldExample valueDublin Core elementRequiredVocabulary or formatNotesRemove

Every catalog has its own field names. One museum calls it “Object name”, another “Title”, a third “Item”. When records move to a search site, a state aggregator or a new collection system, someone has to decide which local field goes where. This worksheet records those decisions once, in writing, so the export, the import and the people checking them all work from the same map.

The target here is Dublin Core, the simplest shared schema and the one most import tools and aggregators understand.

The 15 Dublin Core elements

The Dublin Core Metadata Element Set has fifteen elements. Their names, as DCMI publishes them, are:

Element Plain meaning
title The name given to the item
creator Who made it
subject What it is about
description A summary or note
publisher Who made it available
contributor Others who contributed
date A date in the item’s life, usually creation
type The kind of resource, such as text or image
format File format, medium or extent
identifier An unambiguous reference, such as an accession number
source A resource it was derived from
language The language of the content
relation A related resource
coverage Places or periods it covers
rights Rights held in and over the item

The element set itself does not say which elements a record must have, so your map is where you decide what is required. DCMI encourages implementers to use the more precise DCMI Metadata Terms (the dcterms: namespace), which include fifteen properties with the same names; the mapping decisions are the same.

How to fill it in

  • Your field. The field name exactly as it appears in your current system or spreadsheet.
  • Example value. A real value from a typical record. Examples expose problems that field names hide, such as names written first-name-first or dates written as “Spring ’96”.
  • Dublin Core element. Pick one of the fifteen, or “Not mapped” for fields that should stay internal: insurance values, donor contact details, shelf locations.
  • Required. Whether a record can be published without it. The example requires title, date, identifier and rights.
  • Vocabulary or format. The rule for the value: free text, a date standard, a name authority, a subject list such as Library of Congress Subject Headings (LCSH), or RightsStatements.org.
  • Notes. Anything a person entering or checking data needs to know.

Dates

Write exact dates in ISO 8601 form, year first: 1893-04-12. For approximate or uncertain dates, the Library of Congress’s Extended Date/Time Format (EDTF) adds qualifiers at the end of the date: ~ for approximate and ? for uncertain. “About 1895” becomes 1895~. EDTF’s extended levels are based on features described in ISO 8601-2. Check that your target system accepts EDTF before you rely on it; if it does not, keep a display date (“about 1895”) alongside a sortable year.

Machine suggestions

Dublin Core has no element for where a value came from. If subjects or names were suggested by a model, keep a separate provenance column in the export so reviewed and unreviewed values stay distinguishable. Our AI policy explains how we label machine output until a person has checked it.

Example (fictional). The worksheet’s rows map the Aster County Historical Society’s catalog for letter ACHS-3-012, Ida Okerlund to her sister Marit, April 12, 1893.

  • “Object name” maps to title, with a devised title, “Letter, Ida Okerlund to Marit Okerlund”, because letters carry no title of their own.
  • “Date made” maps to date as 1893-04-12. A cabinet card dated by its mount style would be 1895~.
  • “Writer” maps to creator in inverted form, “Okerlund, Ida”, from a local name authority.
  • “Topics” maps to subject with LCSH headings “Floods; Bridges”. These were machine suggestions and stay flagged until a cataloger reviews them.
  • “Accession number” maps to identifier as ACHS-3-012. The notes ask for the file name stem to be stored in the same record, so the record and its images can be joined. File names that follow a box-and-folder pattern like the file naming builder’s differ from the accession number, so keep both.
  • “Copyright note” maps to rights. With RightsStatements.org as the vocabulary, the value is one statement name exactly as published, here Copyright Not Evaluated, with its web address. That matches the rights review worksheet example, which assigns Copyright Not Evaluated to this letter until the writer’s death date is researched.

Using the map in a project

Agree the map before cataloging starts, not after. Then:

  1. Export a handful of records with the map applied and open them in the target system.
  2. Check every required field arrives, dates sort correctly and names are not split or reversed.
  3. Fix the map, not the records, and run the test again.
  4. Keep the final map with the delivered metadata so anyone can trace a field back to its source.

Our AI cataloging service delivers records as Dublin Core, MARC 21, EAD or CSV for your system’s import, built from a map like this one. The fictional example collection shows finished records with machine and reviewed fields side by side, and the Reading Room’s cataloging and metadata section covers description in more depth.

Send one box first.

A pilot box shows you the images, the recognized text and the catalog records before you commit a whole collection. Tell us what is in the box and what you want people to find.

Test your mapping on a pilot box

Studio and contacts

24 South Weber Street, Suite 225
Colorado Springs, CO 80903
(719) 244-8450sales@dastok.com