Is an adjective always before a noun?

No. English adjectives can occur in predicates, and adjective placement differs across languages.

Are all descriptive words adjectives?

No. Nouns, verbs, and phrases can also describe things. Grammatical category depends on how the expression functions.

Can the same spelling have different grammatical roles?

Yes. A tagger should classify the occurrence in context, using the conventions of its annotation scheme.

AI Glossary / Natural language and speech / Beginner

Adjective

Definition

An adjective is a word class whose members typically express properties of nouns or the things those nouns refer to. In “a quiet room,” quiet modifies room. In “the room is quiet,” quiet functions as a predicate describing the room. Natural language processing systems identify adjectives to help analyze structure, descriptions, comparisons, and opinions in text.

What it means

Adjectives can describe size, color, condition, origin, or evaluation, among other properties. The category is grammatical, not a guarantee that a word expresses an opinion. “Rectangular” can describe shape, whereas “excellent” often expresses an evaluation.

The classification depends on language and annotation scheme. Universal Dependencies uses the tag ADJ, but distinguishes adjectives from determiners, cardinal numbers, and nouns used to modify other nouns. A useful glossary explanation therefore separates the broad linguistic concept from the exact labels a particular dataset assigns.

How it works

A part-of-speech tagger considers the word in context and predicts a grammatical category. A parser can then identify how an adjective relates to a noun or to a larger clause. These are different tasks: recognizing an adjective does not automatically identify everything it describes.

In English, an adjective may appear before a noun, as in “a durable case,” or in a predicate, as in “the case seems durable.” Some adjectives participate in comparisons, such as “smaller” and “smallest,” but not every adjective behaves identically.

The system must also handle ambiguity. A familiar spelling can have different roles in different sentences. Participial forms may behave like verbs or adjectives depending on their use, and different annotation policies can draw the boundary differently. Context and documented conventions are more reliable than a rule that labels every descriptive-looking word an adjective.

Three practical examples

1. Product search: In “lightweight waterproof jacket,” adjective-noun relationships can help extract desired properties. The system still needs to interpret which product is being described and whether those properties are mandatory.

2. Review analysis: “The screen is sharp but the battery is weak” associates different evaluations with different components. Collecting adjectives without their targets loses that distinction.

3. Writing assistance: A tool can highlight repeated adjectives such as “great” for an editor's attention. The grammatical label does not by itself justify removing a word or changing the writer's tone.

Think of it this way

Think of adjectives as property labels attached to items in a catalog: color, condition, size, and similar descriptions. The analogy helps show what many adjectives contribute, but grammatical structure is more flexible than a catalog. A property can be expressed without an adjective, and an adjective can describe something in a full clause rather than sitting directly beside its noun.

Worked example

Take the invented sentence “The small delivery van is surprisingly quiet.” Ignoring punctuation, its seven word tokens are The, small, delivery, van, is, surprisingly, and quiet. Making the token list explicit gives a clear denominator for any later proportion.

Under an English Universal Dependencies-style analysis, small and quiet are adjectives. Small modifies van; quiet describes the van in the predicate. Surprisingly is an adverb modifying quiet. Delivery is a noun used in a compound, not an adjective merely because it appears before van.

There are therefore two adjective tokens among seven word tokens, a proportion of 2/7, or about 28.6 percent. This arithmetic describes the chosen sentence and tokenization convention, not its quality or emotional tone.

Now change the sentence to “The small delivery van is not quiet.” The adjective count remains two and the word count remains seven, but the claim about quietness reverses. An adjective-only summary that records “quiet van” would miss the negation. The example shows why tags, grammatical relations, and sentence meaning need to be considered together.

Illustrative case study

Fictional case: Alder Marketplace builds a tool to group product reviews by discussed features. An initial prototype collects descriptive words and displays “strong” beside every product mentioned in a review.

Reviewers discover that one sentence praises a strong handle while criticizing the case. The team revises the analysis to link descriptions with their targets and preserve negation. The fictional case illustrates an information-extraction problem; it does not claim measured gains or suggest that adjective tagging alone solves review understanding.

Common misconceptions and mistakes

A common mistake is calling every noun modifier an adjective. In “coffee cup,” coffee is commonly analyzed as a noun modifying another noun. Another is assuming that all adjectives convey sentiment: many provide neutral descriptions.

Do not rely on word endings alone. Some English adjective endings are useful clues, but they are not complete rules. Also avoid treating a high adjective count as proof that writing is persuasive, subjective, or exaggerated. Those judgments require context and a defined analysis method.

Business use

Adjective analysis can support catalog enrichment, search, review summarization, and editorial assistance. Its business value usually comes from identifying a property together with its target and qualifications, rather than producing a list of isolated descriptive words.

Teams should document the tagging scheme and inspect errors on their own text. Product names, fragments, abbreviations, and informal reviews may behave differently from well-edited prose. If tags feed a later decision, evaluate the resulting extracted facts as well as token-level tagging accuracy.

Check your understanding

In “The metal box is heavy,” should metal automatically receive the adjective tag because it modifies box?

Answer: No. Under the Universal Dependencies convention, a noun used in a noun compound remains a noun. Heavy is the clear adjective in the predicate. The chosen tagging policy should be stated.

Related terms

Part of speech
Noun
Adverb
Predicate (logic)
Sentiment analysis
predicate
dependency parsing
attributive modifier
comparative

Learn next

Compare part-of-speech tags with dependency relations. Practice distinguishing “what kind of word is this?” from “which expression does this word modify?” Then add negation and coordination to the examples.

Part of speech

Sources

Source: Universal Dependencies, ADJ guidelines, read for adjective uses, noun-modifier distinctions, and scheme-dependent participial cases: https://universaldependencies.org/u/pos/ADJ.html . The supplied full-book URL could not be retrieved. Examples and application discussion are original; language-specific annotation guidelines should govern detailed tagging decisions.

Previous: Adjacent values Next: Adjoint state method