How English Works | The Taxonomy

A library owns ten thousand books.

A shop sells fifty thousand products.

A museum holds a million objects.

A government records thousands of activities.

The problem is not only storing them.

It is deciding what kind of thing each one is.

History?

Art?

Technology?

Education?

Children’s fiction?

Science fiction?

Young adult?

What if one object belongs in several places?

This is taxonomy territory.

A taxonomy is English used to turn a messy domain into named concepts and navigable relationships.

Quick Read

One-sentence answer: a taxonomy works when it gives a domain a controlled set of concepts, labels and broader/narrower relationships so people and systems can classify, browse and retrieve information without pretending that the classification is identical to reality itself.

W3C’s SKOS primer explains how thesauri, classification schemes and other knowledge-organisation systems can be represented as concept schemes. SKOS allows a concept to have a preferred label, alternative labels, broader and narrower relationships, and links to a concept scheme.

Those ideas give us the taxonomy’s central architecture:

concept: the thing being classified at the conceptual level;
preferred label: the main name used in the scheme;
alternative label: synonyms or other acceptable terms;
broader concept: the more general category;
narrower concept: the more specific category;
top concept: an entry point near the top of the hierarchy;
scheme: the bounded classification world in which these relationships operate.

The Taxonomy Is Not the Schema

A schema says what fields a record may contain.

A taxonomy says what conceptual categories can be used to classify something.

Schema:

book_record contains title, author, subject and publication_date.

Taxonomy:

Science → Biology → Ecology

The schema creates the slot called subject.

The taxonomy can supply the controlled concepts allowed inside that slot.

schema = shape of record
taxonomy = organisation of concepts used by the record

Preferred Labels Reduce Synonym Chaos

One person writes:

AI

Another:

artificial intelligence

Another:

machine intelligence

A controlled vocabulary can select one preferred label while preserving alternatives as synonyms or alternate access terms.

SKOS explicitly provides preferred and alternative lexical labels for this purpose.

one concept can have several names without becoming several concepts

This distinction is essential for search and indexing.

Readers can search using familiar language while the system stores a stable conceptual identity.

Broader and Narrower Are Meaning Relationships

SKOS uses broader and narrower relationships to connect more general and more specific concepts.

Animals → Mammals → Cats

“Mammals” is narrower than “Animals.”

“Animals” is broader than “Mammals.”

This hierarchy supports browsing.

A user who does not know the exact term “Felidae” may begin at Animals and travel downward.

taxonomy is search by conceptual neighbourhood

Broader Does Not Automatically Mean Part Of

Consider:

Vehicle → Car → Wheel

Something has gone wrong.

A wheel is part of a car.

It is not a kind of car.

The W3C SKOS primer uses a closely related example to warn that broader/narrower hierarchies are not automatically transitive in every messy knowledge-organisation system.

This matters because taxonomies can collapse several relation types into one tree.

kind-of, part-of and instance-of are different semantic relations even when a crude hierarchy makes them look alike

Classification becomes more trustworthy when the relation itself is named accurately.

One Concept Can Have More Than One Parent

W3C notes that a SKOS concept can have several broader concepts at the same time.

Consider:

Dog → Mammal
Dog → Domesticated Animal

Both routes can be useful.

This breaks the idea that every taxonomy must be one perfect tree.

Real domains often have several legitimate dimensions.

A book can be:

historical fiction;
young adult;
Singapore literature;
war literature.

Forcing it into only one path may simplify navigation while destroying useful meaning.

multiple inheritance is what happens when reality has more than one useful axis

Top Concepts Are Entry Points, Not Ultimate Truths

SKOS concept schemes can identify top concepts—general entry points into the scheme.

A museum might use:

Art
History
Science and Technology
Everyday Life

These top concepts shape how users enter the collection.

Another museum could choose completely different top concepts for the same objects.

top level is an editorial decision about the first useful distinctions

Taxonomy therefore contains perspective even before the user clicks anything.

A Concept Scheme Defines a Local Classification World

SKOS represents controlled vocabularies through concept schemes.

This is important because the same concept can participate in different schemes.

“Cats” in a general animal scheme may have different neighbouring concepts from “Cats” in a veterinary scheme or pet-store taxonomy.

same concept, different scheme, different useful neighbourhood

Classification meaning is therefore partly local to the purpose of the scheme.

Preferred Labels Are Language-Specific

SKOS allows preferred labels to carry language tags.

A concept can have one preferred English label and another preferred French label.

This helps multilingual classification avoid a false choice between:

one universal label

and:

different concepts for every language

The concept can remain stable while labels vary by language.

This is especially important in global knowledge systems where translation is an access layer rather than a change of identity.

Categories Have Edge Cases

Where does a documentary-drama belong?

Film?

Documentary?

Drama?

History?

What about a smartwatch?

Watch?

Computer?

Health device?

Communication device?

Taxonomies become most interesting at the edges because the boundary reveals what principle the classification is using.

the difficult object is often the best test of the category definition

Granularity Determines How Fine the World Becomes

A taxonomy can classify books as:

Fiction → Fantasy

Or:

Fiction → Fantasy → Urban Fantasy → Singapore Urban Fantasy

Which is better?

It depends on the job.

Too coarse, and useful distinctions disappear.

Too fine, and users face hundreds of almost-empty categories.

granularity is the resolution at which the taxonomy chooses to see

A good taxonomy uses enough detail to support retrieval and analysis without creating complexity that no real user needs.

“Other” Is a Diagnostic Category

Every taxonomy eventually meets something that does not fit.

Many systems create:

Other

This can be useful as a pressure-release valve.

It can also become a graveyard.

If 35% of records land in Other, the taxonomy may be missing an important conceptual branch.

Other is not merely a bucket; it is evidence about what the classification failed to anticipate

Good taxonomy maintenance watches the size and composition of miscellaneous categories.

Taxonomies Drift Because the World and Its Language Change

New technologies appear.

Old categories become offensive or obsolete.

Disciplines reorganise.

Industries merge.

Words change meaning.

A taxonomy that never changes eventually describes the history of its designers more accurately than the domain it is meant to classify.

taxonomy version = classification decisions at a particular moment in knowledge

Review therefore needs to consider both new concepts and old relationships that no longer make sense.

Renaming a Category Is Not Always a Semantic Change

A concept may keep the same identity while its preferred label changes.

This is one reason SKOS distinguishes concepts from lexical labels.

Old label:

Electronic Learning

New preferred label:

Digital Learning

Perhaps the conceptual category stayed largely stable.

In another case, a renamed category may hide a genuine conceptual change.

label change and concept change should not be assumed to be the same event

Reclassification Changes History Without Changing the Past

An object classified as “Computing” in 1995 may later be reclassified under “Mobile Communication.”

The object itself did not change.

The classification model changed.

This creates important provenance questions:

What was the old category?
When did reclassification occur?
Which taxonomy version made the new decision?
Can historical searches reconstruct the old classification?

classification history is not object history, but both can matter

Taxonomies Shape Search Results

A filter saying:

Subject: Science

depends on a taxonomy upstream.

Which concepts count as Science?

Does Environmental Studies sit under Science, Geography or both?

A search system can look neutral while inheriting every classification choice made earlier.

taxonomy → metadata assignment → filter membership → what the user sees

This connects directly to Filter and Search Query. Search does not begin only when the user types. It begins when the collection was classified.

Classification Can Carry Institutional Bias

Who created the categories?

Whose language became the preferred label?

Which distinctions were considered important?

Which identities were grouped together?

Which were given their own category?

Classification is never merely alphabetical housekeeping.

A taxonomy can reproduce the assumptions of the institution that built it.

The categories through which a system sees the world can become the categories through which the organisation acts on the world.

This does not mean classification is impossible.

It means classification deserves review, provenance and the ability to change.

Taxonomy and Facets Solve Different Problems

A single hierarchy asks:

Where does this belong?

Faceted classification asks several questions at once:

What subject?
What age group?
What format?
What location?
What time period?

This can be more faithful for objects that have several independent dimensions.

The taxonomy can supply controlled concepts inside each facet.

hierarchy gives one route; facets give several coordinate axes

Primary School: Sort the Same Objects in Two Ways

Give students twelve classroom objects.

First classify by material:

paper;
plastic;
metal;
wood.

Then classify by use:

writing;
measuring;
storing;
cutting.

The same object can move because the classification question changed.

The lesson is simple and profound:

categories depend on the distinction you are trying to make

Lower Secondary: Find the Broken Hierarchy

Give students:

Transport → Cars → Wheels

Ask why the hierarchy feels wrong.

Then compare:

Transport → Road Vehicles → Cars

and:

Car → has part → Wheel

This teaches semantic relation types rather than merely tree drawing.

Upper Secondary: Audit a Classification World

What is the purpose of the taxonomy?
What are the top concepts?
Which labels are preferred?
Which synonyms are preserved?
Are broader/narrower relations really kind-of relations?
Can concepts have multiple parents?
What repeatedly falls into Other?
Is the granularity useful?
How has the taxonomy changed over time?
Whose assumptions are embedded in the categories?

This turns taxonomy reading into a lesson in vocabulary, information architecture, power and epistemology.

Ten Failure Modes of Taxonomy English

Synonym fragmentation. Several labels accidentally become separate categories for one concept.
Label-concept confusion. A wording change is mistaken for a conceptual change, or vice versa.
Relation collapse. kind-of, part-of and instance-of are treated as one generic hierarchy.
Single-parent forcing. Multi-dimensional concepts are squeezed into one artificial branch.
Granularity mismatch. Categories are too broad to retrieve usefully or too narrow to manage.
Other overload. Miscellaneous becomes the largest category because the scheme stopped evolving.
Top-level bias. The first distinctions reflect organisational habit rather than user need.
Taxonomy drift. New concepts and changed language accumulate without controlled revision.
Reclassification amnesia. Categories change but the historical classification path disappears.
Institutional blindness. Important realities remain hard to see because the taxonomy never created a category for them.

How to Build a Better Taxonomy

What domain and user job does the scheme serve?
What are the most useful top concepts?
Which concepts need stable identities?
What preferred and alternative labels are required?
Which relationships are genuinely broader/narrower?
Where is multiple inheritance legitimate?
What granularity supports retrieval?
How will Other and edge cases be reviewed?
How will changes and reclassifications be versioned?
What assumptions or blind spots should be audited?

Then perform the difficult-object test:

Take the object that fits least comfortably. Does the taxonomy explain why it belongs where it does—or does the object reveal that the classification principle was never clear?

The Deeper Idea: Taxonomies Are Maps of Difference

Reality does not arrive with folder names attached.

Humans create distinctions.

They name them.

They arrange them.

They use those arrangements to search, count, compare and decide.

continuous world → chosen distinctions → named concepts → classification → retrieval and action

This is why taxonomy belongs to how English works.

It is language used not merely to describe the world, but to decide which differences will become navigable.

Reader Checklist

What is this taxonomy for?
What are the top concepts?
Which labels are preferred?
What synonyms exist?
What relations connect concepts?
Can concepts belong to several broader categories?
What lands in Other?
Is the granularity useful?
Which version applies?
What does the taxonomy make easy—or hard—to notice?

Research and Further Reading

W3C — SKOS Simple Knowledge Organization System Primer

Final idea: a taxonomy is good when it makes a domain easier to navigate without pretending the map of categories is the same thing as the world being classified.