New independent field guide, published 17 October 2026. Examples are illustrative. The historical directory records linked below are separate source context, not current recommendations or earlier versions of this essay.
A category tree looks like a neutral arrangement of folders. In practice it is a set of decisions about which distinctions matter and where people should look first. A resource about historic maps can belong to geography, history, education or a regional collection, depending on the task. Forcing one answer can make the tree tidy while making the resource harder to find. This guide develops a small, maintainable classification scheme using invented examples. The aim is not a universal taxonomy. It is a structure whose categories have clear meanings, whose relationships can be explained and whose changes do not lose the resources already described.
Choose the question a category answers
Before naming branches, decide whether they describe subject, format, audience, location or another dimension. A label such as history describes a subject; video describes a format; beginners describes an audience. Placing all three as equivalent siblings can work for a small browsing menu, but it creates ambiguity about what membership means. A beginner's history video then seems to require three competing homes. In an invented research collection, keep subject categories separate from optional format and audience facets so that each label answers a distinct, predictable question about the resource.
Facets do not require a complicated interactive application. A static resource page can display subject, format and geographic coverage as labelled fields, with links to suitable index pages. What matters is conceptual consistency. An editor should not have to guess whether a region refers to the publisher's address or the material covered by the resource. Define that distinction in the field description. If both are useful, store them separately. A structured record is valuable because it prevents the same short label from silently carrying several incompatible meanings across the collection.
A concept needs more than a preferred spelling
W3C's SKOS model distinguishes concepts from the words used to label them and provides ways to express preferred and alternative labels. The distinction is useful even in a handwritten taxonomy. An invented concept for public transport can retain one stable identifier while accepting transit as an alternative search term. Changing the display wording does not require treating the subject as a new concept. A stable identifier also lets descriptions and review history remain attached when language is improved. The label is the reader-facing expression, not the entire identity of the category.
Write a short scope note for ambiguous concepts. A category called design might include graphic, industrial and interaction design, or it might intentionally cover only one of those areas. The note should say what belongs and what nearby subjects do not. Use a few positive and negative examples rather than a long abstract definition that editors cannot apply. In a fictional directory, an article about designing a scientific experiment might be excluded from visual design while remaining discoverable under research methods. That decision becomes consistent only when the intended boundary has been stated.
Use hierarchy for a real broader relationship
A broader category should genuinely contain or encompass the narrower subject in the sense defined by the scheme. It should not simply mean that two topics are often mentioned together. In an invented collection, astronomy can be a narrower subject within science, while telescopes may have a related link to observing practice. The related relationship is useful without pretending that every observing-method resource is a type of telescope. Mixing association with hierarchy makes the tree harder to reason about and can generate strange inherited groupings when the collection grows.
Avoid designing every resource into an unnecessarily deep path. A chain of five nearly empty categories can create more decisions than information. Start with the distinctions users actually need and add depth when the material justifies it. Conversely, one giant miscellaneous branch can hide important differences. The right balance is tested through retrieval tasks, not through symmetry on a diagram. Ask a reader to find a plausible resource without telling them the intended path. Their first category choice can reveal whether the labels match the language and mental model the collection is trying to serve.
Allow multiple routes without duplicating the record
A resource can be linked from several relevant category pages while retaining one underlying record. This is different from copying its description into independent entries that later diverge. An invented oral-history archive might appear under regional history and recorded interviews, with both links leading to the same record. Its URL, description and review events then have one maintenance location. The category pages provide different routes to it. This approach separates the problem of discovery from the problem of identity, reducing the cost of a later correction or destination change.
Cross-links should have a reason. Adding every vaguely related category to every record recreates an unfiltered search space inside the taxonomy. A useful test is whether the resource substantially helps answer a question associated with the category, rather than merely mentioning its name. For a fictional map lesson, cartography might be a strong subject relationship while the location of the author's office is not. Record the relationship's meaning when needed. Subject coverage, publisher location and intended audience can all support discovery, but they should not be disguised as interchangeable kinds of relevance.
Plan for renaming, merging and splitting
Taxonomies change as language and collections change. Renaming a category can preserve its identifier and update the display label while keeping an old label as a lookup aid. Merging two categories requires a decision about their records and old routes. Splitting a broad category requires actual reassessment; it cannot be completed simply by creating two new empty folders. Keep a change note explaining the decision and its effective date. That note lets future editors understand why an older record used a label that no longer appears in the current navigation.
A historical taxonomy needs a different treatment from a current one. An old category spelling or count is evidence about the saved directory, not necessarily an error to repair silently. This edition's archive retains captured category labels as metadata and does not claim they are today's best classification. If a new reading guide uses improved terminology, label it as new writing. Preserving the old and designing the new are compatible when their roles are separate. The trouble begins when a modern reorganisation is presented as though it were the original site's exact structure and historical wording.
Test the structure with small retrieval exercises
Create a few invented tasks that reflect the collection's purpose: find a beginner-friendly explanation, locate a resource covering a region, or compare two formats for the same subject. Try them through category browsing and through whatever search method is available. Note the wrong turns and the labels people expected to see. A task can reveal missing alternative terms, confusing facet names or a category that combines unrelated ideas. It can also reveal when a resource description is the problem rather than the taxonomy itself. Do not rebuild the entire tree in response to every single hesitation.
Maintain a compact category register with identifier, preferred label, alternatives, scope note, broader relationships and change history. That is enough to make many small collections more coherent without implementing all of SKOS. The historical Arts and Regional records linked below show captured classification choices from the old directory, not an endorsed modern taxonomy. The central principle is to make each relationship mean something explicit. A reader should be able to arrive by a sensible route, recognise where they are and change direction without being punished for having framed the subject differently from the editor.