A concept identifier is a stable code or reference used to identify a particular concept independently of the human-readable label displayed to readers. A library may show the term Equivalent Fractions today and later adopt a clearer display label, while continuing to use the same internal identifier for the same scoped idea. Other records can refer to that identifier without having to guess whether the latest wording is the same subject as before.
This is a practical reason vocabulary can become a technology. When words work as addresses, a change of spelling, language or preferred title should not accidentally break links between articles, lessons, sources and searches. Stable identifiers help organisations avoid duplicates, recognise renamed concepts, connect language variants, manage corrections and explain whether two things really are identical or merely related. But an ID is not a magic guarantee of truth: it must refer to a carefully defined concept and be maintained responsibly.
This guide answers what concept identifiers are, why labels change, how identifiers are assigned and preserved, what should happen when a concept splits or merges, and how language, versioning and canonical links interact. It draws on W3C guidance for stable Web identifiers and SKOS concept schemes, with original school, vocabulary and knowledge-library cases. The educational examples are fictional, and no particular code presented here is an official external subject or curriculum identifier.
Editorial note: The concept IDs, learning examples and local library scenarios are fictional. A stable ID is an organisational tool, not proof of an official designation or a verified educational claim.
Contents: concepts, names, stable IDs and identity changes
Open 48 chapters, worked cases and exercises
- Begin with a title that changes
- A label is for reading; an identifier is for reference
- Why spellings should not create duplicate topics
- An identifier need not be a readable English word
- A URI can serve as an identifier
- A permalink and a concept ID are different things
- Why W3C recommends stable URIs
- Hash and 303 URI strategies
- Concept scope should be documented
- An ID is not proof of factual accuracy
- The same label can refer to two concepts
- Different labels can refer to the same concept
- Preferred labels can change without changing identity
- Alternative labels are not alternative identifiers
- A retired label can remain searchable
- An identifier must not be reused casually
- A concept split needs special care
- A concept merge is not just deleting one row
- A concept can change gradually
- Names may carry dates and version boundaries
- The difference between a concept and an instance
- A person identity needs a different governance model
- Canonical sources and concept identifiers
- Concept IDs enable structured relations
- Concept identifiers and controlled vocabulary
- Concept identifiers and faceted classification
- ID collisions can create hidden errors
- A code should have a documented namespace
- Readable IDs versus opaque IDs
- Slugs are convenient but may be unstable
- A crosswalk links identities across schemes
- Exact matching is stronger than close matching
- Do not misuse owl:sameAs
- An old ID can lead to a current entry
- Source and editorial history belong with the record
- What happens when an official term changes?
- A small namespace and concept-table example
- How to test the identity after a rename
- How to diagnose an accidental duplicate
- How to diagnose a false merger
- A concept identifier cannot replace reader navigation
- A concept identifier cannot settle a disputed definition
- A 20-minute classroom exercise
- What a sound governance rule looks like
- Five diagnostic mistakes
- Ten quick questions
- Frequently asked questions
- Research notes and next step
1. Begin with a title that changes
Imagine a local learning library originally labels a collection ‘Fractions that look different but mean the same amount’. The editor later chooses the preferred label ‘Equivalent Fractions’ to make the topic easier to browse.
If the scope remains genuinely the same, the underlying concept can keep its identifier. A reader following a historic bookmark or an article reference can still reach the intended idea even though the displayed wording changed.
The key question is whether this is only a change in label or a change in what the category means. An identifier should preserve the same intended referent, not hide a silent change in scope.
2. A label is for reading; an identifier is for reference
A label is a human-friendly name, often chosen for clarity. An identifier allows records, software or editors to refer to the concept consistently without relying entirely on the current label.
The W3C SKOS Reference models concepts as identifiable resources with preferred, alternative and hidden labels. The concept’s identity is separate from the strings used to name it.
A small school may use a local ID in a spreadsheet rather than deploy a Semantic Web system. The distinction remains useful, although a local code has no automatic global meaning.
3. Why spellings should not create duplicate topics
A student searches ‘Maths vocabulary’ and another searches ‘Mathematics vocabulary’. If the library gives each phrase an independent concept identity, articles can be split across two unnecessary collections.
An editor can determine whether the expressions denote the same scoped subject, then use one stable concept identifier with appropriately managed alternative labels.
Do not merge two different meanings merely because their spelling is close. The identity decision depends on what the topic covers, not the editor’s desire for a shorter list.
4. An identifier need not be a readable English word
A fictional vocabulary scheme might use IDs such as VOC-001 and VOC-002. These codes are useful only inside the environment that assigns and manages them.
The public label may be readable and translated, while the internal ID remains unchanged. A person can discover the concept through natural language and a system can use the ID to maintain relationships.
Never mistake a made-up example code for a real qualification code, official subject ID or permanent Internet identifier.
5. A URI can serve as an identifier
On the Web, a concept can be identified by an IRI or URI, which may allow different systems to refer to the same resource. The W3C document Cool URIs for the Semantic Web explains design approaches and the importance of stable, manageable identifiers.
The URI can lead people or software to descriptions of the concept without making the human-readable label its only identity.
Making a URI stable and resolvable requires real hosting, redirects, metadata and long-term maintenance. Merely writing a URL-shaped string does not build that infrastructure.
6. A permalink and a concept ID are different things
A blog post URL identifies a document or resource route. A concept identifier is intended to identify the subject concept represented by one or more documents.
One article can discuss several concepts, and a single concept may be taught in many articles. Equating every article slug to a concept would make editorial changes harder.
Use distinct identifiers when the distinction matters. A page can display and explain a concept without becoming identical to that abstract concept.
7. Why W3C recommends stable URIs
Web identifiers are valuable when they remain usable over time and are easy to manage. The Cool URIs guidance emphasises simplicity, stability and manageability.
Changing hosting technologies, themes or page slugs should not require renaming a concept. A lasting identifier design avoids tying the identifier to replaceable implementation details where feasible.
This is an architectural principle rather than a guarantee that every URI remains online forever. Organisational responsibility and redirects still matter.
8. Hash and 303 URI strategies
The W3C Cool URIs note describes hash URIs and HTTP 303 redirects as ways to distinguish non-information resources from documents that describe them in some Linked Data deployment patterns.
A vocabulary project does not necessarily need either approach for an internal ten-row worksheet. Large public cross-system schemes may have additional requirements for dereferencing and content negotiation.
Choose the implementation according to actual use and technical capacity; do not claim that a WordPress permalink automatically satisfies every Semantic Web deployment requirement.
9. Concept scope should be documented
An ID cannot protect meaning if editors cannot tell what it refers to. Record a plain-language definition, scope note, examples and a plausible exclusion for each consequential concept.
Suppose an entry called ‘Reading’ could mean decoding print, comprehending fiction or reading the instructions for a Mathematics problem. Its intended scope must be specific enough for users to classify material consistently.
The scope note does not have to be encyclopaedic. It needs enough detail to distinguish neighbouring concepts that would otherwise be confused.
10. An ID is not proof of factual accuracy
A database item can have an elegant identifier and still contain an inaccurate explanation. The identifier supports tracking and retrieval, not independent verification of its claim.
If an article about evaporation misstates a scientific condition, assigning it a permanent ID will not correct the science. An editor still needs to review the underlying content.
Keep identity checks and factual checks separate. Both are useful, but they solve different problems.
11. The same label can refer to two concepts
‘Scale’ can describe a ratio, a musical organisation or a measuring instrument. A system storing resources from several school subjects should not assume one identity follows from one word.
Concept IDs can distinguish the different senses while labels and subject contexts help the reader choose the relevant one.
A useful test asks whether two records could substitute for one another in a student’s actual question. If not, the shared spelling may hide different concepts.
12. Different labels can refer to the same concept
A topic may be labelled in English, Mandarin or another language. Different readers can use different phrases while the scheme refers to one appropriately scoped concept.
However, language equivalence is not guaranteed by machine translation. A term in another language may carry different boundaries or educational conventions.
Consult the Lexical Gap guide before asserting that a translation is an exact match.
13. Preferred labels can change without changing identity
A library might replace a confusing title with clearer wording while retaining the same subject definition. The old expression can remain an alias for search.
The change should be recorded with an effective date or editorial note if earlier content depends on it. Readers then understand why a saved search or older article uses a different term.
Do not quietly rewrite historic quoted terminology as though an external institution used the new name at the earlier date.
14. Alternative labels are not alternative identifiers
An alternative label is another expression for finding or displaying a concept; an alternate identifier is another code or reference that a system treats as identifying it under some mapping policy.
The two categories behave differently. Adding a colloquial phrase does not by itself establish that another organisation’s code identifies the same concept.
An editor should record where a term came from and separately assess any cross-system identity claim.
15. A retired label can remain searchable
An older school resource collection might have a former local name that families still search for. Keeping it as a historical alias can help users find the updated description.
The public page should state when a real official naming rule changed and what remains different; a redirect must not imply that all policies are unchanged.
A retired label is a navigation aid. It is not an instruction to revive outdated information as current fact.
16. An identifier must not be reused casually
If VOC-001 originally identifies a fractions concept, later reassigning VOC-001 to a science topic would make historic links and records misleading.
A robust governance policy avoids reusing identifiers for different meanings. New concepts should receive new IDs, while retired IDs remain documented as retired where appropriate.
The policy is simple in principle but requires discipline when editors delete or reorganise collections.
17. A concept split needs special care
Suppose one broad concept ‘Vocabulary Practice’ has become too vague and is split into ‘Retrieval Practice’ and ‘Vocabulary Use in Writing’. These are not merely two new names for precisely the same previous concept.
Create appropriate new identities for the genuinely distinct scopes. Decide whether the old concept remains as a broad parent, redirects to a choice page or is retired with a history note.
Do not quietly attach the old identifier to just one child if that would misrepresent all records previously classified under the broader meaning.
18. A concept merge is not just deleting one row
Two near-duplicate concepts may genuinely cover the same idea. The editor can consider merging them after comparing definitions, examples, user intent and existing links.
A merge should preserve a trace of replaced IDs and update or redirect references where feasible. Otherwise historic records may appear to vanish.
Near-synonymy is not sufficient justification. A merge that combines different curriculum stages or requirements can damage accuracy.
19. A concept can change gradually
Over time, a research area may adopt new terminology while retaining much of its earlier meaning. Another area may shift its boundary so substantially that continuity of identity becomes contested.
Document what changed in scope and why the same ID was retained or a new one created. There may not be one universally correct answer for all conceptual evolution.
The important practical quality is a reviewable decision rather than an invisible assumption that names always track reality perfectly.
20. Names may carry dates and version boundaries
An official educational programme may have a particular name and syllabus at one date and revised requirements later. Those periods need careful documentation.
A concept ID used for general browsing should not conceal assessment changes that matter to students. Some datasets may need separate version-specific records as well as a broader programme identity.
Use current authoritative sources for claims about eligibility, assessment and curriculum rather than rely on a stable label.
21. The difference between a concept and an instance
Fractions is a general mathematical concept; one student’s answer ‘3/4’ in a particular exercise is an instance or data item in a task.
An ID identifying the concept ‘Equivalent Fractions’ should not be confused with the identifier of that individual worksheet answer.
This distinction helps when building a graph or database: resource type, subject concept and a particular event can each require their own identity.
22. A person identity needs a different governance model
Names of real people can be duplicated, changed or used in different forms. Identifying people across systems carries privacy and safety responsibilities that differ from indexing Mathematics topics.
A school learning vocabulary should not casually publish personal identifiers or aggregate private learner records because stable IDs are technically convenient.
Use concept identifiers for shared educational subjects and apply legitimate privacy controls to any individual data.
23. Canonical sources and concept identifiers
A concept record may link to an authoritative definition, a local teaching article and additional explanatory sources. Those resources support the record without becoming identical to the concept itself.
Record the source and date for externally defined terminology, especially when a policy or technical standard is involved.
An ID makes it easier to keep evidence attached to the right concept, while a provenance record shows where the explanation came from.
24. Concept IDs enable structured relations
A knowledge graph can say that a resource teaches a concept, that one concept is related to another, or that a term has a narrower concept under a scheme.
Stable IDs make those links more resilient than repeatedly matching human-readable text strings. However, the relations must be explicitly defined and checked.
See Knowledge Graphs for the separate question of modelling entities and relationships.
25. Concept identifiers and controlled vocabulary
A Controlled Vocabulary manages labels, scope notes and concept relations. Stable identifiers give those managed concepts persistent reference points.
The vocabulary can start in a simple table; the concept identifier is one field among several. Having the field does not by itself mean the scheme’s terminology is consistent.
Use both the ID and definition when reviewing duplicate terms. The code cannot decide semantic equivalence for you.
26. Concept identifiers and faceted classification
A document may be classified by subject, stage, task type and format. Each facet value may come from a carefully managed set of concepts.
Stable IDs help distinguish subject ‘Reading Comprehension’ from resource type ‘Comprehension Worksheet’ even if the display labels share a word.
The Faceted Classification article explores why independent axes should not be collapsed into one oversized title.
27. ID collisions can create hidden errors
Two editors working independently might both assign the same local number to different new concepts. Combining their spreadsheets can silently attach articles to the wrong entry.
A central allocation rule, unique namespace or appropriate generated identifier can reduce this risk. The chosen method depends on the size and technical design of the system.
Before importing records, test whether any identifier is duplicated with incompatible definitions. An attractive label does not protect against an ID collision.
28. A code should have a documented namespace
A local string such as ‘001’ is ambiguous outside its table. If several subject collections independently use ‘001’, a shared warehouse needs a way to tell which scheme issued which code.
A namespace or URI design can place the code within a recognised issuing context. This is part of making identifiers interoperable rather than merely unique within one spreadsheet.
Do not confuse a globally unique-looking code with a globally recognised authority. The relevant governance and resolution mechanism still matter.
29. Readable IDs versus opaque IDs
A readable ID may encode a mnemonic such as MATH-EQF-001. This can help editors inspect records, but it may become misleading if the concept moves to another subject grouping.
An opaque ID avoids tying meaning to particular letters, but humans need labels and documentation to interpret it. Both approaches have trade-offs.
Choose based on expected stability and workflows. Do not force an ID change simply because an editor has renamed a category.
30. Slugs are convenient but may be unstable
A webpage slug such as ‘what-is-fractions’ is human-readable but may be edited for clearer search intent. If the slug is treated as the only concept identifier, the change can break references.
A stable internal ID and appropriate redirects can help manage page renames. The article permalink and concept ID still serve different purposes.
Changing a slug does not guarantee that old external links automatically redirect. Verify the site’s actual behaviour.
31. A crosswalk links identities across schemes
One organisation’s concept for ‘Secondary Mathematics’ may have a different scope from another system’s similarly named entry. A crosswalk records a reasoned relationship between the two.
The IDs should remain distinct until the mapping has been examined. A mapping does not necessarily justify collapsing them into one global identity.
The forthcoming Semantic Crosswalk article develops these relation types and the risks of asserting equivalence too broadly.
32. Exact matching is stronger than close matching
The W3C SKOS mapping section distinguishes skos:exactMatch and skos:closeMatch along with broader, narrower and related mappings.
Exact matching indicates high confidence that two concepts can be used interchangeably across a wide range of information-retrieval applications. Close matching indicates sufficiently similar meanings for some retrieval uses, and is not declared transitive.
These are statements about concept correspondence, not automatically an assertion that the resources are the same Web individual under owl:sameAs.
33. Do not misuse owl:sameAs
Formal ontology identity claims such as owl:sameAs are stronger than a typical close correspondence between two subject concepts. The SKOS reference explicitly distinguishes them.
An educational crosswalk that says two stages have related learning content should not use such a strong identity claim unless its strict meaning is actually justified.
Use the weakest accurate relationship that supports the desired task. False equivalence is often more harmful than a deliberately qualified link.
34. An old ID can lead to a current entry
An editor may decide to retire an erroneous duplicate while preserving its old ID as a redirect or replacement reference to a surviving record.
The mapping should state why the replacement is considered valid and what happened to content classified under the retired identifier.
A redirect is a technical operation, not proof that the two historic definitions were identical. Keep an appropriate change note.
35. Source and editorial history belong with the record
A useful concept register can contain identifier, preferred label, definition, source, creation date, review decision and a link to the current authoritative record.
When something changes, record the previous form and the reason. That history can help resolve questions arising from articles written under older labels.
The Provenance Trail guide explores how to track origin and transformation beyond a single concept record.
36. What happens when an official term changes?
An external standard may rename a concept without preserving every previous condition. A local vocabulary should verify what changed before updating the preferred label or mapping.
If the scope genuinely changed, the team may need a new record or explicit version distinction rather than one silent label replacement.
This is especially important for public educational advice, where the current source should be linked and dated.
37. A small namespace and concept-table example
Create a fictional table with columns Concept ID, Preferred Label, Scope Note, Alternative Labels and Source. Use three rows such as EDU-001 for ‘Equivalent Fractions’, EDU-002 for ‘Comparing Fractions’ and EDU-003 for ‘Simplifying Fractions’.
These are made-up local IDs. The concepts overlap but perform different learning tasks, so separate entries may be useful. A query can navigate between them through related or narrower links.
Ask a second editor whether the scope notes make the differences clear. If not, improve the definitions before expanding the identifier list.
38. How to test the identity after a rename
Give a learner two versions of a concept page with different labels. Ask whether both versions describe the same idea and support the same example and non-example.
If yes under the agreed scope, preserving identity may be reasonable. If one version now includes new situations that the other excluded, further review is needed.
The test is qualitative evidence about scope, not an automated theorem that a shared heading proves identity.
39. How to diagnose an accidental duplicate
Two concept IDs may have near-identical labels. Compare their definitions, content examples, exclusions and existing links.
Do not immediately delete one record. Some apparent duplicates encode genuinely different stages, languages, curricular interpretations or versions.
Only after reviewing the relationship should the editor decide to merge, map, rename or keep both. The decision should be recorded.
40. How to diagnose a false merger
A library puts ‘Fractions’ and ‘Percentages’ under one concept ID because both use rational-number ideas. Their relationship is real, but their roles in a lesson differ.
Separate the concepts where the indexing task needs the distinction, then use a relevant relationship or cross-topic explanation instead of pretending they are synonyms.
The repair improves retrieval precision and preserves what students need to learn.
41. A concept identifier cannot replace reader navigation
An internal code EDU-047 may be stable but meaningless to parents. The website still needs plain-English labels, helpful summaries and navigation that begins with real questions.
Use the code behind the scenes to keep links consistent; show a suitable title and examples to the reader.
This is the division of labour between machine identity and audience design. One cannot substitute for the other.
42. A concept identifier cannot settle a disputed definition
Two teachers can disagree about the boundary of a local term while both pointing to the same ID. The identifier records which entry they are discussing, not which definition they ought to choose.
Review the intended meaning, evidence and governance authority. If the definition changes, document whether the concept identity remains appropriate.
The Metalinguistic Negotiation guide explores the difference between disagreement about facts and disagreement over which meanings to adopt.
43. A 20-minute classroom exercise
Begin with five small topic cards, two of which use alternative labels for the same idea and two of which have similar-looking but different meanings. Ask students to propose identifiers and preferred labels.
Introduce a rename and a false synonym. Students must decide when to preserve an ID, when to add an alias and when a new ID is appropriate.
Finish by asking another pupil to search the resulting table using natural language. The task measures whether the scheme supports the intended retrieval, not whether the codes look sophisticated.
44. What a sound governance rule looks like
One concise policy could say: ‘Every concept receives a stable local ID. Editors may change labels after review, but must not reuse an ID for an unrelated concept; scope changes require a documented decision.’
Add responsibilities for reviewing synonyms, redirects and external mappings when the library grows. A smaller team may have one editor; a public multi-organisation system needs stronger coordination.
Governance is part of the technology. The code itself cannot prevent an authorised but careless editor from making a damaging change.
45. Five diagnostic mistakes
One: changing a concept ID every time the title changes. Two: treating two identical labels as necessarily one subject. Three: mapping two neighbouring concepts with a stronger identity relation than the evidence supports.
Four: using a blog URL as if it were the same as a subject concept. Five: assigning the same ID to different objects after a merge or import.
Each mistake calls for identity review, clearer scope notes or a repair to the identifier system. None is solved merely by making the ID longer.
46. Ten quick questions
1. Is an identifier the same thing as a label? No. 2. Can a label change while a concept remains stable? Yes. 3. Can two concepts share a displayed word form? Yes. 4. Can one concept have multiple language labels? Yes.
5. Does a stable ID prove that a claim is true? No. 6. Should an ID be reused for an unrelated idea? Generally no in a persistent scheme. 7. Can a split require new IDs? Yes. 8. Can a mapping be close but not exact? Yes.
9. Is a post permalink identical to the abstract concept it describes? Not necessarily. 10. What is the purpose? Maintain an inspectable identity while names, documents and systems evolve.
47. Frequently asked questions
Is this relevant to children’s vocabulary? Yes, as a way to understand why names, categories and labels can change while the underlying concept remains identifiable. Do I need special software? A small reviewed table can teach the principle.
Should every word have a global ID? Not for ordinary language learning. IDs are useful where records must be connected and governed consistently. What if meanings overlap? Record an appropriate relation instead of assuming identity.
Why is it a vocabulary technology? It makes concepts referable across revisions and languages while allowing people to keep using natural words.
48. Research notes and next step
The W3C SKOS Reference explains identifiable concept resources, labels and concept-mapping properties. Cool URIs for the Semantic Web describes stability and manageability of Web identifiers. The Dublin Core Metadata Terms provides further guidance on resource descriptions and identifiers.
All example EDU codes and classroom records here are invented. The public standards do not validate an unverified local concept or establish equivalence between real school systems.
Choose one concept used across several pages. Write its scope, preferred label and current references. Ask what would break if the title were changed next year, and how a stable identifier and redirect or mapping policy could preserve access. That question shows why a concept ID is more than a decorative code.