Can one word belong to more than one category?

A spelling can occur in different categories depending on its use, as with noun and verb uses of record.

Is noun the same as subject?

No. Noun is a category; subject is a grammatical function that a phrase may perform.

Are all tag sets identical?

No. They differ in granularity and conventions, so labels should be interpreted within the chosen scheme.

AI Glossary / Graphs and symbolic AI / Beginner

Lexical category

Definition

A lexical category is a grammatical class associated with a word or lexical item, such as noun, verb, adjective, or adverb. In many NLP contexts, the term is used broadly for word classes or parts of speech. Some linguistic frameworks use it more narrowly for content-word categories, so the intended terminology should be stated.

What it means

Categories describe how words behave in grammatical structures, not simply what topics they refer to. A noun is not adequately defined as “a thing,” nor a verb as “an action,” because abstract nouns and state-denoting verbs also occur.
The category assigned to a particular occurrence depends on its use. The same spelling can function differently across sentences. For example, record is a noun in “Keep a record” and a verb in “Record the result.” A language-processing system therefore needs context rather than a dictionary that assigns every spelling one permanent grammatical label.

How it works

A part-of-speech tagger analyzes words in context and assigns labels from a defined tag set. It may use word forms, neighboring words, grammatical patterns, and learned contextual representations. Different tag sets distinguish categories at different levels of detail.
Category information can help parsing, information extraction, and other language tasks, but it is only one layer of analysis. Features such as tense or number may be represented separately, as may syntactic relationships such as subject and object. A category label also does not resolve every word sense: two different meanings can both be nouns or both be verbs.

Three practical examples

A parser distinguishes a verb use of record from a noun use before constructing a sentence structure.
A terminology tool looks for noun phrases as candidate concepts while checking that the extracted phrases make sense in context.
A grammar-analysis exercise compares an adjective modifying a noun with an adverb modifying a verb or adjective.

Think of it this way

Think of a word taking a role in a particular scene. The same actor can play different roles in different productions. The analogy is limited because grammatical categories follow language-specific distributions and annotation conventions, rather than being freely assigned by a director or determined solely by a word's meaning.

Worked example

Consider two invented sentences: “They record results” and “The record is complete.” In the first, record functions as the main verb. In the second, record is the head noun in the phrase “the record.” The spelling is identical, but its grammatical behavior differs.
Now consider “The light box moved” and “They light the lamp.” Under the intended first reading, light is an adjective describing the box as not heavy. In the second sentence, light is a verb. If “light box” instead names a type of illuminated device, the analysis may require a different treatment of the compound. The intended reading and annotation scheme matter.
Suppose an invented tagger processes 100 reviewed occurrences of record. Seventy are nouns and thirty are verbs. A simplistic rule that always labels record as a noun achieves 70 percent accuracy but misses every verb occurrence. A contextual tagger correctly labels 65 nouns and 25 verbs, for 90 correct out of 100, or 90 percent accuracy.
The comparison illustrates why context and category-specific errors matter. It does not claim that these two sentences determine a complete tagging model or that a particular real tagger has the stated performance.

Illustrative case study

In a fictional archive project, Birch Terms extracts product concepts by treating every occurrence of common dictionary nouns as a term. It mistakenly extracts words that are functioning as verbs in the documents.
The team adds contextual grammatical analysis and reviews the resulting phrases against the intended terminology task. It also documents how its tag set handles compounds and proper names. The invented case illustrates the relationship between grammatical labeling and an application; it does not claim that part-of-speech tagging alone can produce a correct domain glossary.

Common misconceptions and mistakes

A lexical category is not the same as a syntactic function. A noun phrase may serve as a subject or object, while noun is a category of its head word. Likewise, tense and number are grammatical features rather than necessarily separate major word classes.
Do not assume that every language uses English categories in exactly the same way, or that every annotation project draws identical boundaries. The phrase lexical category itself can contrast with functional category in some theories. Finally, correct category tagging does not guarantee correct meaning interpretation, factual accuracy, or a complete sentence analysis.

Business use

Category information is useful when a business NLP task depends on grammatical patterns. Examples include extracting candidate terms, identifying descriptions, or interpreting short commands. Choose a tag set whose distinctions support the task rather than adding labels merely because they are available.
Evaluate the system on the actual domain, including abbreviations, product names, fragments, and unusual uses of familiar words. Examine errors by category and by ambiguous spelling. Keep grammatical labels separate from the final business meaning assigned to an expression. This makes it easier to see whether an error arose during tagging, structural analysis, or later interpretation.

Check your understanding

Why is the always-noun rule inadequate in the worked example despite scoring 70 percent?

Answer: The dataset contains many noun occurrences, so the aggregate score hides complete failure on the thirty verb occurrences. It does not use context to distinguish the two categories.

Related terms

Part of speech
Syntactic categories
Noun phrase
Lexical ambiguity
POS tagging
syntactic category
functional category
grammatical feature
parser

Learn next

Collect several sentences in which the same spelling functions differently. Assign categories using one documented tag set, then separate those labels from sentence functions and word senses. Note any cases where the intended reading or annotation convention changes the answer.

Sources

Sources: Universal Dependencies, official Universal POS tags reference: https://universaldependencies.org/u/pos/ Jurafsky and Martin, Speech and Language Processing, archived chapter on word classes and sequence labeling: https://web.stanford.edu/~jurafsky/slp3/old_aug24/17.pdf

Previous: Lexical ambiguity Next: Lexical choice