Lexical Semantics in SEO: Word Relations Search Engines Use

LEXICAL RELATIONS

Lexical semantics is the study of what words mean and how word meanings relate to each other, and in SEO it is the set of word relations a search engine uses to connect queries, pages and entities. Six lexical relations matter: synonymy, antonymy, hypernymy, hyponymy, meronymy and polysemy. Lexical search matches the words of a query with the words of a page, semantic search matches their meanings, and modern search engines combine both. Search engines use lexical relations to expand queries, classify pages and resolve ambiguous words. In a topical map, hypernyms and hyponyms set the page hierarchy, meronyms become attributes, synonyms are held to one name per concept, ambiguous words are resolved in the first sentence, antonyms shape contrast sections, headings and anchors carry the relations, and a five-step check reviews any page. Holistic Radar’s own pages follow the same rules, and the questions at the end cover LSI keywords and synonym lists.

What lexical semantics is

Lexical semantics is the branch of linguistics that studies the meaning of words and the relations between those meanings. It asks what a word refers to, how many senses it has, and which other words it is a kind of, a part of, the same as or the opposite of.

kind ofkind ofsame aspart ofCarconceptVehiclehypernymSedanhyponymAutomobilesynonymEnginemeronym
Figure 1 · One concept, four word relations

Search engines need those relations because searchers and writers rarely use the same words. A searcher types “car”, a page says “vehicle”, another says “sedan”, and a fourth says “automobile”. Matching them requires knowing that sedan is a kind of car, that automobile means car, and that a vehicle includes cars and much else. Lexical semantics supplies that knowledge.

Lexical semantics differs from the wider semantics of a page. Lexical semantics works at the level of single words and their relations; the semantics of a page also covers sentences, facts and the order in which they appear. A page can use every word correctly and still state its facts badly, so lexical rules are a floor, not the whole method.

The best-known machine-readable source of lexical relations is WordNet, a lexical database of English started at Princeton University in the 1980s under the psychologist George A. Miller. WordNet groups words into sets of synonyms and links the sets by relations such as “kind of” and “part of”, the same structure a topical map uses for attributes and pages.

The six lexical relations

Six lexical relations account for most of what a search engine needs from word meaning.

RelationDefinitionExample
SynonymyTwo words with the same meaning in a contextCar and automobile
AntonymyTwo words with opposite meaningsOrganic and paid
HypernymyA word for the broader classVehicle is a hypernym of car
HyponymyA word for a kind within the classSedan is a hyponym of car
MeronymyA word for a part of the wholeEngine is a meronym of car
PolysemyOne word with several related senses“Crawl” as a swimming stroke and as a bot visiting pages
01Synonymysame meaning02Antonymyopposite meaning03Hypernymythe broader class04Hyponymya kind within it05Meronymya part of the whole06Polysemyone word, several senses
Figure 2 · Six relations a search engine reads

Homonymy is the limit case of polysemy, where one spelling carries unrelated meanings, such as “bank” for a river edge and a lender. For a search engine both create the same problem: the word alone does not say which meaning the page uses.

Lexical search matches the words of a query with the words of a document, and semantic search matches their meanings. A lexical system ranks a page for “affordable sedan” when the page contains those words, typically weighting rare words above common ones with a formula such as BM25. A semantic system represents the query and the page as vectors and ranks a page about “low-cost four-door cars” for the same query, because the meanings sit close together.

DimensionLexical searchSemantic search
What it matchesWords and their frequenciesMeanings, as vectors
StrengthExact terms, names, codes, rare wordsSynonyms, paraphrases, intent
WeaknessMisses pages that use other wordsCan blur exact names and numbers
Page that winsThe page that uses the query’s wordsThe page that covers the query’s meaning
WORDSLexical searchQuery words matched to page wordsRare words weighted, e.g. BM25Wins on names, codes, numbersMEANINGSSemantic searchQuery and page as vectorsClose meanings matchWins on synonyms and paraphrasevsSearch engines combine both, so a page needs exact names and full coverage
Figure 3 · Words matched or meanings matched

Modern search engines combine the two, which is why a page needs both: the exact name of each entity and attribute, for lexical matching, and complete coverage of the meaning, for semantic matching. The semantic search article covers how Google moved from one to the other.

How search engines use lexical relations

Search engines use lexical relations in three places: before retrieval, during classification and during disambiguation.

  1. Query expansion. Before retrieval, a query is widened with synonyms and close variants, so a search for “car insurance quote” also retrieves pages that say “auto insurance quotes”.
  2. Classification. Hypernyms place a page in a class: a page about sedans, hatchbacks and SUVs is classified as a page about cars, even if it never uses the word.
  3. Disambiguation. Surrounding words resolve polysemy: “crawl budget” beside “Googlebot” is about bots, and “front crawl” beside “breathing” is about swimming.
Query expansionsynonyms widen the queryClassificationhypernyms place the pageDisambiguationcontext picks the sense
Figure 4 · Three uses, three stages

Each use rewards a page that makes its word relations explicit. A page that names the class, lists its kinds and parts, and places its ambiguous words beside the words that resolve them gives a search engine fewer guesses to make.

Hypernyms and hyponyms in a topical map

Hypernyms and hyponyms set the hierarchy of a topical map. The hypernym names the page that covers the class, and each hyponym names a page, or a section, that covers one kind within it. A page about SEO audits is the hypernym page; technical audits, content audits and semantic SEO audits are its hyponyms.

SEO auditshypernym pageTechnical auditshyponymContent auditshyponymSemantic SEO auditshyponymEvery hyponym page links up; the hypernym page lists every hyponym
Figure 5 · Hypernym page above, hyponym pages below

The relation decides two things. It decides the heading order, because a page names the class before its kinds, and it decides the links, because every hyponym page links up to its hypernym and the hypernym page lists its hyponyms. A searcher who lands on a hyponym page and wants the broader view is one link from it, and a search engine reading the hypernym page sees the whole class in one place.

Hypernyms are also how a page states its type in its first sentence. “A content brief is a document that…” places the content brief in the class of documents, and the rest of the sentence names what separates it from other documents.

Meronyms as attributes

Meronyms, the parts of a whole, become the attributes of an entity in a topical map. A topical map has a central entity, a source context, a core section, an outer section and bridges; each part is a meronym of the topical map and a candidate for its own heading or page.

Central entitypartSource contextpartCore sectionpartOuter sectionpartBridgespartTopical mapthe wholeEach part is a candidate heading or page
Figure 6 · The parts of a topical map

The part-of relation also tests completeness. A page about a topical map that never names one of its parts is missing an attribute the consensus expects, and an entity gap list will show it. Listing the parts of the whole, then giving each one a heading, is the simplest way to make a page cover what its subject contains.

Synonyms: one name per concept

Synonyms are where SEO practice and lexical semantics disagree most often. Older practice rotated synonyms to avoid repetition; semantic SEO gives each concept one name and uses it every time. A site that calls one concept “topical map” on one page, “topic map” on another and “content map” on a third splits the signal for that concept across three strings.

SPLITSynonym rotationtopical map on page Atopic map on page Bcontent map on page CCLEANOne name per conceptSynonym once, in the definingsentenceOne name in headings and anchorsOne name in structured datavs
Figure 7 · Rotation splits the signal

Search engines already know the synonyms. Query expansion connects “topic map” to “topical map” without the page using both, so rotation adds nothing the engine needs and costs consistency. Synonyms still have a place inside a sentence that defines a term, where they help a reader recognize it: “a topical map, sometimes called a topic map, is…”. After that sentence, the page uses one name.

The rule extends across the site. The same concept carries the same name in headings, anchors, structured data and every article, which is what keeps the site’s n-grams clean.

Polysemy and disambiguation

A polysemous word is resolved in the first sentence that uses it. The resolution is a short qualifier that names the sense: “crawl, the visit a search engine bot makes to a URL”, or “authority, a site’s demonstrated coverage of a subject”. Once the sense is stated, the rest of the page can use the word alone.

Disambiguation matters most for entities that share a name. A page about Mercury the planet, Mercury the element and Mercury the record label needs to state which one it covers before it states anything about it, because every fact on the page attaches to the entity the search engine resolves. A fact attached to the wrong entity is not a weak signal; it is a wrong one.

Antonyms and contrast sections

Antonyms shape the contrast sections searchers expect on definitional pages. A page about organic search is expected to say how it differs from paid search; a page about topical authority, how it differs from domain authority. The opposite term defines the subject by its border.

A contrast section states both terms, the dimension they differ on, and the value each takes. A table works well because each row is one dimension and each column one term. The section stays short when the opposite has its own page: it states the difference and links there.

Lexical relations in headings and anchors

Headings and anchors are where lexical relations are read most directly, because each is short and each names one thing. A heading sequence that moves from hypernym to hyponym, from “SEO audits” to “semantic SEO audits”, tells a reader and a search engine that the page narrows as it goes. A sequence that jumps between classes, from “SEO audits” to “link building” to “audit tools”, signals a page without a hierarchy.

Anchors follow the one-name rule. An internal link to the topical map page uses “topical map” as its anchor, not “this guide” or “our mapping article”, so the anchor carries the same name the target uses in its title. Across many links, consistent anchors are one of the clearest statements a site makes about which page owns which concept.

A lexical check for one page

A lexical check reviews one page in five steps:

  1. Find the first sentence that defines the subject and confirm it names the hypernym.
  2. List the headings and confirm they move from the class to its kinds and parts.
  3. List every name used for the page’s main concept and reduce them to one.
  4. Find each polysemous word and confirm its sense is stated where it first appears.
  5. Confirm the contrast section names the opposite term the consensus expects, and links to its page.
Hypernymin thedefinitionHeadingsclass tokindsOne nameper conceptSensesresolvedearlyContrastoppositenamed
Figure 8 · Five steps per page

The check takes minutes per page and catches most of the word-level problems that make a page harder to classify.

Lexical semantics on holisticradar.com

Holistic Radar applies the same relations to its own pages. Semantic SEO is the central entity, and its hypernym, SEO discipline, appears in the first sentence that defines it. The articles at /learn/ are hyponyms and meronyms of Semantic SEO: kinds of analysis, parts of a topical map, stages of retrieval.

Each concept has one name across the site. “Topical map” is never “topic map” in a heading, and “query network” is never “keyword universe”. The audit that checks these names also finds the exceptions: “cost of retrieval” appears 148 times across the site and “retrieval cost” 20 times, a split the audit flags for correction. Contrast sections follow antonyms the consensus expects: topical authority against domain authority, topical map against topic cluster, semantic search against keyword search.

The lexical rules are the sentence-level half of the method; microsemantics covers the rest, from modality to numeric precision.

Lexical semantics questions

Are LSI keywords the same as lexical semantics?

LSI keywords are not lexical semantics. Latent semantic indexing is a retrieval technique described in 1990 for finding term co-occurrence in small, closed document sets, and “LSI keywords” is a marketing label for related words. Lexical semantics describes the relations between word meanings, which is what search engines actually use to connect queries and pages. The semantic keywords article covers the difference in detail.

Should a page include a list of synonyms?

A page should not include a list of synonyms. A synonym helps once, inside the sentence that defines a term; a list of synonyms adds strings without meaning and splits the page’s name for its subject.

Does lexical search still matter?

Lexical search still matters for exact names, product codes, numbers and rare terms, where meaning-based matching can blur. A page that states the exact name of each entity serves both systems.

Lexical Semantics in the semantic SEO topical map

Lexical Semantics is the word-relation attribute of the page semantics entity inside the Holistic Radar topical map, and it sits in the outer section under the Know predicate.

The outer section holds the page because it explains how word meaning connects queries and pages without selling a service, and its bridge passes session signals toward Semantic Copy, where every client page is written with one name per concept.

→ Semantic copy: client pages written with one name per concept and explicit word relations.

→ What are microsemantics: the sentence rules beside the word rules.

→ What is semantic search: how search moved from words to meanings.

→ Semantic keywords: why LSI keywords are the wrong model.

Get my free Radar ScanFree Radar Scan