A semantic crosswalk is a documented mapping between concepts, terms or categories in two information systems. It helps readers and software connect a label used in one library with the closest appropriate concept used in another, while preserving differences in scope. If one study collection uses ‘Fraction Equivalence’ and another uses ‘Equivalent Fractions’, an editor can examine whether they describe sufficiently corresponding ideas. The mapping is a reasoned relationship, not merely the discovery that two titles look alike.
This is vocabulary as interoperability technology. Parents, students, tutors and research readers encounter knowledge organised by different publishers, schools, countries and disciplines. A crosswalk can make relevant materials discoverable across those boundaries without pretending that teaching stages, assessment rules or concept definitions are identical. A good mapping says whether the concepts are equivalent for retrieval, merely close, broader, narrower or related—and records what evidence justifies the choice.
This guide explains what a semantic crosswalk is, why it differs from translation and simple synonym matching, how W3C SKOS mapping relations work, and how to avoid false equivalence. We use original, explicitly hypothetical education-library cases, a practical mapping register, collision checks, examples of version drift and a quality review that tests both correct and incorrect matches. The result should be a navigable library that can connect useful knowledge without silently changing its meaning.
Editorial note: School subjects and collections mentioned in the worked cases are invented mappings for teaching. They should not be mistaken for official equivalences, admissions advice or verified alignments between external curricula.
Contents: map concepts without losing meaning
Open all 46 sections and worked mappings
- Two libraries, two ways of naming one subject
- The formal foundation in SKOS
- A crosswalk records relationships, not title replacements
- Start with stable concept identifiers
- Define the source scheme and target scheme
- Exact match is a strong retrieval claim
- Close match is deliberately weaker
- Broad match preserves different scope
- Narrow match is the inverse direction
- Related match is not hierarchy
- Avoid simultaneously contradictory relations
- Exact match is not owl:sameAs
- Language translation and concept mapping differ
- Curriculum correspondence is not an admissions decision
- Map the intended learning outcome
- Concept scope notes are essential evidence
- Check examples and non-examples
- A mapping can depend on the retrieval application
- Why multi-hop mapping can drift
- Transitivity makes exact matches high-stakes
- A mapping is not a copy of someone else’s vocabulary
- Worked case: two names for the same Mathematics idea
- Worked case: stage-specific vocabulary topics
- Worked case: reading inference and background knowledge
- Worked case: Science topics with shared terms
- Worked case: a changed standard
- Mappings need a provenance trail
- A compact mapping register
- Proposed, reviewed and published are different states
- Where false merges cause practical harm
- Where missing mappings waste time
- A crosswalk is not query expansion
- A crosswalk is not a knowledge graph by itself
- What to do when there is no match
- What to do when several targets look suitable
- Human review and automation have different roles
- Crosswalk quality requires negative tests
- A published link should be rechecked after changes
- Cross-language review can require specialist help
- A mapping should not determine a pupil’s eligibility
- A four-step reader-facing explanation
- A workshop with three fictional collections
- Diagnostic: six mapping mistakes
- Ten questions and short answers
- Frequently asked questions
- Primary sources and your next action
1. Two libraries, two ways of naming one subject
Imagine Library A uses the preferred concept label ‘Equivalent Fractions’. Library B has a concept called ‘Equal Fraction Values’, defined through examples showing that two fractions can represent the same number. A parent searching across both may want both collections.
The editor cannot choose an exact mapping solely from the two labels. They inspect definitions, examples, excluded meanings and relevant use. If those agree closely enough for the stated retrieval application, the crosswalk can record a justified correspondence.
This is a relationship between scoped concepts in separate systems. It is not an instruction to rename Library B or claim that the institutions have agreed to the same curriculum.
2. The formal foundation in SKOS
The W3C SKOS Reference, Mapping Properties defines five main mapping relations: skos:exactMatch, skos:closeMatch, skos:broadMatch, skos:narrowMatch and skos:relatedMatch.
SKOS is a model for knowledge organisation systems, including thesauri and concept schemes. The SKOS Primer provides worked explanations of modelling and mapping.
An informal school spreadsheet can use similar categories in plain language. It must not claim to implement SKOS formally unless its identifiers, relations and data actually conform to that model.
3. A crosswalk records relationships, not title replacements
Library A may have a broad topic ‘Reading Comprehension’, while Library B distinguishes inference, reference and evidence questions. Replacing all B categories with A’s heading would erase useful differences.
A mapping can preserve the broader relation while allowing readers to discover narrower guides through an appropriate search or navigation interface.
The job is to preserve each library’s meaning and make the connection visible. One dictionary of labels cannot substitute for this scope analysis.
4. Start with stable concept identifiers
A concept identifier provides a persistent point of reference even when a human-readable label changes. The mapping record should identify the source scheme and concept, the target scheme and concept, and the intended relation.
If a WordPress page title changes, the concept mapping should not automatically become a new mapping unless the intended subject changed. Conversely, a major scope change may require review even if the visible title remains identical.
Read the Concept Identifier guide for the difference between names and persistent concept identity.
5. Define the source scheme and target scheme
A crosswalk needs to say which vocabulary supplied each concept. A simple code such as 042 is ambiguous if two libraries assign the same number.
Use source scheme name or namespace, concept ID, current label and suitable version information. The target side needs equivalent fields.
Do not treat a page URL, a subject identifier and a qualification code as interchangeable simply because all appear as strings in the same spreadsheet.
6. Exact match is a strong retrieval claim
SKOS skos:exactMatch expresses high confidence that two concepts can be used interchangeably across a wide range of information-retrieval applications. It is not a casual statement that two words have similar connotations.
The standard makes exact match symmetric and transitive. Two successive exact mappings can lead to a further inferred exact mapping, increasing the consequences of any erroneous assertion.
Before choosing exact match, compare the full scope and typical queries. Use a weaker relation if meaningful differences remain.
7. Close match is deliberately weaker
SKOS skos:closeMatch means concepts are sufficiently similar to be used interchangeably in some information-retrieval applications. It does not promise all uses or all contexts.
The W3C standard does not declare close match transitive. If A is close to B, and B close to C, it does not logically follow under SKOS that A is close to C.
This matters for educational crosswalks across several providers: a chain of individually reasonable approximations can drift far from the original meaning.
8. Broad match preserves different scope
If one scheme identifies ‘Quadratic Equations’ and another has ‘Algebra’, the former may be narrower than the latter under an appropriately checked classification.
A skos:broadMatch from the narrower concept to the broader one can express the hierarchical mapping. The direction matters.
A broad match is not exact match. A learner asking specifically for quadratic equations should not be sent indiscriminately to every algebra resource and told the intent was fully satisfied.
9. Narrow match is the inverse direction
SKOS skos:narrowMatch links a broader source concept with a narrower target concept. Under the standard’s property definitions, it is the inverse of broad match.
Mapping direction should be displayed in readable language beside the identifiers, such as ‘Source is broader than target’ or ‘Source is narrower than target’.
This prevents editors from swapping the relation simply because the two records appear in a different column order.
10. Related match is not hierarchy
Vocabulary acquisition and reading comprehension are often related educational topics, but one need not always be a subtype of the other. In a suitable pair of schemes, an associative mapping may be more accurate.
SKOS skos:relatedMatch represents such an associative mapping between concepts. It should not be used to conceal an uncertain exact correspondence.
Ask what relation an actual reader needs. ‘Related’ may guide further reading without implying substitution.
11. Avoid simultaneously contradictory relations
The SKOS reference states that exact match is disjoint with broader-match and related-match relationships to the same concept. An entry asserting both exact and hierarchical or associative mapping for the same pair can violate the SKOS integrity model.
A spreadsheet validator should flag multiple incompatible relation choices rather than quietly accept them all.
The purpose is not formalism for its own sake. Conflicting labels can make retrieval rules unpredictable and obscure which correspondence an editor actually believes.
12. Exact match is not owl:sameAs
SKOS notes that owl:sameAs denotes identity of individuals in OWL, whereas skos:exactMatch concerns concept correspondence suitable for information retrieval.
Mapping two related subject terms should not automatically collapse them into the same Web individual or assert formal ontology identity. That is a stronger and different claim.
Use the relation that expresses the evidence and intended application. An unnecessarily strong identity statement can propagate errors across connected datasets.
13. Language translation and concept mapping differ
A word in English and its translated counterpart may seem equivalent, but languages sometimes divide concepts differently or use terms with different register and scope.
First check the concept in each scheme; then record its labels in the relevant languages. A translation is not automatically a proven exact mapping.
The Lexical Gap guide shows why word-to-word matching may require explanation rather than a single substitution.
14. Curriculum correspondence is not an admissions decision
Two school systems may teach a similar algebraic idea at different years. That similarity can support topic navigation, but it does not establish grade, qualification or assessment equivalence.
Official educational policy questions require verification against the relevant current awarding body and institution. A crosswalk should state when its mapping is only conceptual.
Do not convert a close topic match into a claim that one course automatically satisfies another course’s entry requirements.
15. Map the intended learning outcome
A broad article about fractions may explain meanings, comparison, operations and applications. A narrower worksheet may test only equivalent fractions.
To assess mapping, identify the learning outcome the source record serves. A similarity in vocabulary does not guarantee equal pedagogical function.
A useful crosswalk can state ‘Related for navigation; not interchangeable for task assessment’ where that describes the actual relationship.
16. Concept scope notes are essential evidence
Preferred labels alone are often too short for a sound mapping decision. Read the source scope note and the target definition alongside examples and explicit exclusions.
If neither scheme has clear documentation, mark the mapping provisional or unresolved rather than inventing equivalence.
A transparent uncertainty state is often more useful than a confident false link that will be reused across hundreds of pages.
17. Check examples and non-examples
Create one example that both concepts should include and one case that one includes but the other excludes. The difference can reveal that exact match is too strong.
Consider a source concept encompassing all algebraic expressions and a target concept limited to linear expressions. A quadratic example belongs to one but not the other.
This is a direct way to test scope using school material instead of relying only on abstract definitions.
18. A mapping can depend on the retrieval application
Two concepts might be similar enough that users searching broadly for introduction material benefit from seeing both. They may still be too different for a strict syllabus-alignment table.
Close match can express this application-limited usefulness. The crosswalk record should state the intended use and any excluded uses.
Without purpose, ‘close’ becomes a vague judgement whose practical consequences cannot be evaluated.
19. Why multi-hop mapping can drift
Library A’s concept is a specialised topic. Library B uses a slightly broader label, and Library C uses an even broader interpretation. Pairwise similarity does not guarantee the first and third remain close.
This is one reason SKOS closeMatch is not transitive. A system should not automatically infer A–C close correspondence from two close links.
When a user follows several crosswalks, show the chain or re-evaluate the final pair directly before presenting equivalence.
20. Transitivity makes exact matches high-stakes
In SKOS, exact match is transitive. If A exact-matches B and B exact-matches C, then A exact-matches C follows under the model.
One mistaken exact mapping can therefore contaminate a wider network of inferred connections. A review process should apply stronger standards to high-impact exact-match assertions.
Use verified scope descriptions, tested examples and appropriate provenance. The formal rule increases the value of quality control.
21. A mapping is not a copy of someone else’s vocabulary
Different institutions may have good reasons to use different classifications. A crosswalk connects them without requiring either to surrender its preferred labels or internal structure.
That property is valuable in a library-of-libraries setting where readers want to navigate across credible sources without erasing the original publishers’ meaning.
Keep source identity and target identity visible. A crosswalk should not present a locally invented label as the official source’s own terminology.
22. Worked case: two names for the same Mathematics idea
Fictional Library A stores a concept ‘Equivalent Fractions’ with examples 1/2 = 2/4 and 3/6 = 1/2. Library B uses ‘Equal Fractions’ with an explicit definition saying two written fractions have the same numeric value.
After reviewing the definitions and exclusions, an editor may judge them equivalent for the intended study-material retrieval. The exact relation is an editorial assertion that needs the scheme records as evidence.
Now imagine Library B also includes a separate meaning involving equal numbers of fractional objects. The broader interpretation could undermine exact mapping. Check the actual scope rather than the headline.
23. Worked case: stage-specific vocabulary topics
Library A defines ‘Primary 5 Vocabulary for Writing’ and Library B defines ‘Upper Primary Vocabulary’. The second entry includes reading, speaking and writing across several years.
The concepts overlap but are not interchangeable. Depending on the precise scheme definitions, A may be narrower than B or may need a qualified related mapping.
Do not quietly rename A as B to fill a missing link. The narrower learning stage and writing purpose matter to the parent searching.
24. Worked case: reading inference and background knowledge
One library uses ‘Reading Inference’ broadly, covering predictions, bridging and interpretation. Another maintains a distinct concept for ‘Bridging Inference’.
The latter may fit beneath the broader reading-inference concept under an appropriately documented educational taxonomy.
Read the Bridging Inference guide to see why one subskill should not be mistaken for every form of inference.
25. Worked case: Science topics with shared terms
Two indexes contain ‘Change of State’. One concerns phases of matter; another concerns workflow states in a digital document system. Their labels look alike but denote different subjects.
A crosswalk mapping them exactly would be false. The correct action may be to keep the concepts separate and use context-sensitive labels.
Homographs and familiar shared words are common collision sources. Concept definitions must decide identity, not surface resemblance.
26. Worked case: a changed standard
An external curriculum provider revises one topic definition. The local crosswalk was based on the previous version and has not been reviewed since.
The editor should mark the mapping as requiring reassessment and compare old and new descriptions. The mapping may stay valid, become close rather than exact, or need replacement.
Do not let the unchanged external label create a false impression that its meaning and scope have not changed.
27. Mappings need a provenance trail
For a consequential mapping, record who proposed it, what definitions were compared, when the assessment occurred and which source versions were used.
If a problem appears later, the editor can identify the decision and repair the appropriate link without overwriting unrelated content.
The Provenance Trail guide explains why origin and revision history matter.
28. A compact mapping register
Use fields for Source Scheme, Source ID, Source Label, Target Scheme, Target ID, Target Label, Relation, Scope Evidence, Reviewer, Review Date and Mapping Status.
A small project can begin with fewer fields, but should retain enough to distinguish a verified mapping from a suggestion awaiting review.
Do not claim the invented worksheet implements every SKOS technical requirement. It is a teaching and editorial aid for inspectable decisions.
29. Proposed, reviewed and published are different states
A proposed mapping may be promising while still awaiting a second reader’s check. Publishing it as exact equivalence before review risks exposing users to a false connection.
Keep stages such as Draft, Needs Evidence, Reviewed, Approved and Superseded, aligned with the project’s real approval procedure.
An accurate status can stop a false sense of completion from spreading through an editorial team.
30. Where false merges cause practical harm
If two different Mathematics subject stages are merged, a student might retrieve exercises that assume concepts they have not studied. If two different Science processes are collapsed, a reader may receive an explanation for the wrong phenomenon.
The problem is not simply a slightly untidy index. It can change the learning action the reader takes.
A crosswalk should test realistic user questions and detect results that look plausible but target the wrong scope.
31. Where missing mappings waste time
A reader searches for a familiar abbreviation while the target library uses a complete formal title. Without an appropriate alias or crosswalk, relevant content might not be discovered.
An editor can create a justified path to the target concept without writing a duplicate article or replacing a legitimate official name.
Measure whether the intended material becomes findable for the search phrase. The result depends on the actual search implementation.
32. A crosswalk is not query expansion
A crosswalk records relationships between concepts in distinct schemes. Query expansion modifies or augments a search so it can retrieve records using alternative expressions.
The two can work together: a query might use a verified crosswalk to search an external scheme’s label. Yet query expansion may also use synonyms, spelling correction or user feedback without any formal crosswalk.
Keep the mapping evidence separate from search-ranking behaviour. A well-mapped concept does not automatically make a website’s search engine use it.
33. A crosswalk is not a knowledge graph by itself
A small table containing source IDs, target IDs and typed relations is structured relationship data. It can become part of a knowledge graph when represented and connected appropriately.
A broad knowledge graph may also model articles, authors, sources, evidence and learning prerequisites beyond concept correspondence.
See Knowledge Graphs for the different task of representing and questioning networks of facts.
34. What to do when there is no match
Some source concepts have no appropriate target counterpart. The crosswalk should allow a No Match or Unresolved status rather than forcing a nearest-looking label.
An unfamiliar specialised concept may need a descriptive explanation, a new local concept or a navigation link to the original source.
The absence of an exact match is not a failure when honest non-equivalence protects the reader from a false claim.
35. What to do when several targets look suitable
A broad source concept might overlap several narrower target concepts. This can require multiple mappings with distinct relationships rather than one arbitrary winner.
Record why the broad source cannot be represented by any single target without loss of scope. A search interface may display a choice or a collection of relevant targets.
Do not label a set of narrower targets as one exact match unless the mapping model and actual meaning justify that aggregation.
36. Human review and automation have different roles
Software can propose candidate mappings from labels, definitions or language models. Human editors or validated processes still need to assess whether the relation captures the intended meaning.
A high similarity score is not a proof of conceptual equivalence. Some misleading near matches share most of their words but differ in a crucial negation, condition or audience.
Use automatic suggestions as proposals, keep the source evidence visible and evaluate errors before publishing consequential mappings.
37. Crosswalk quality requires negative tests
A test suite should include a known same-concept case, a clearly broader case, a related-but-not-equivalent pair and an intentionally misleading label collision.
If the system maps ‘evaporation’ exactly to ‘boiling’, the failure reveals that lexical proximity was mistaken for identity.
Good quality assurance checks what must not map, not only whether plausible matches can be found.
38. A published link should be rechecked after changes
An external scheme can move URLs, retire records or alter definitions. A crosswalk may remain conceptually useful while its destination becomes broken or stale.
Review high-impact references and maintain redirects or replacement notes where legitimate. Record the source version used for a previous mapping decision.
Do not delete historic evidence merely because a target has changed. The change itself may explain why the old mapping is no longer valid.
39. Cross-language review can require specialist help
One language’s word for a school stage or subject may not match another system’s category boundaries. A professional or knowledgeable reviewer may need to inspect the actual language use.
A machine-translated label can help discover candidate concepts but does not by itself establish exact equivalence.
Keep labels, translations and mapped concept relationships separate in the record.
40. A mapping should not determine a pupil’s eligibility
A crosswalk may help a family browse similar Mathematics topics in different educational systems. It should not be used as a substitute for official admissions, placement, assessment or qualification decisions.
Such decisions require current authoritative rules that can be more specific than topic similarity. The mapping should clearly indicate its informational purpose.
A misleading educational equivalence can affect real choices, so a careful crosswalk states its limits prominently.
41. A four-step reader-facing explanation
When showing a crosswalk to a reader, state the source concept, the target concept, the meaning of the relation and what the mapping does not prove.
For example: ‘This topic is related to the other library’s Fraction Comparisons collection; the resources cover overlapping ideas but are not interchangeable assessment units.’
A transparent relation is more useful than a short mysterious link labelled ‘Equivalent’ when the underlying scope is different.
42. A workshop with three fictional collections
Create a local Mathematics scheme, a reading-skills scheme and a broader learning-library scheme. Give each four scoped concepts with stable IDs.
Have learners propose one exact, one close, one broader and one related mapping across schemes. Require an example and non-example for each proposed relation.
Then introduce a revised target definition. Students must revisit the mapping and explain whether its status should change. The activity teaches meaning preservation across systems rather than memorising property names.
43. Diagnostic: six mapping mistakes
One: selecting exact match because two labels have the same spelling. Two: treating close match as transitive. Three: reversing broader and narrower directions.
Four: using owl:sameAs when only information-retrieval similarity is justified. Five: presenting a topic crosswalk as official curriculum equivalence. Six: keeping a mapping without source versions or a correction path.
Each error needs a different repair. Deleting every questionable link is not always necessary; some should be downgraded, clarified or reviewed.
44. Ten questions and short answers
1. Is a crosswalk just a translation dictionary? No. 2. Does SKOS support exactMatch and closeMatch? Yes. 3. Is closeMatch transitive? No. 4. Is exactMatch transitive under SKOS? Yes.
5. Does relatedMatch mean one topic is a subtype of another? No. 6. Is a mapping the same as an official course equivalence? No. 7. Can one source have several target relations? Yes, if justified.
8. Should a crosswalk include source versions? It is a useful governance practice. 9. Can no match be a legitimate result? Yes. 10. What is the central purpose? To connect concepts across schemes without losing their intended meaning.
45. Frequently asked questions
Who needs a semantic crosswalk? Libraries, curriculum indexes, research collections and any team connecting differently labelled knowledge systems can benefit. Can a small tutor build one? Yes, start with a reviewed table before introducing complex tooling.
Does exact match mean all linked articles are equally good? No; it concerns concept correspondence, not factual quality or teaching effectiveness. Can mapping improve search? It can support retrieval when the search implementation uses the relations.
What if two systems disagree about scope? Record a qualified relation or no match rather than inventing equivalence.
46. Primary sources and your next action
The W3C SKOS Mapping Properties is the central standard for exact, close, hierarchical and associative concept mappings. The SKOS Primer explains their use with examples. The W3C PROV-O specification supplies a separate model for provenance that can help document mapping decisions.
These sources describe Web standards and information organisation. All school cases, library identities and proposed mapping decisions here are fictional explanations, not verified mappings between actual education authorities.
Select one concept in each of two libraries. Compare the definitions, examples and exclusions. Record the source and target IDs, a justified relation and one circumstance in which the mapping could mislead. Then ask another reader to use the link. A semantic crosswalk is working when it makes knowledge easier to navigate without inventing equality.