AI Glossary / Natural language and speech / Beginner

Determiner

Definition

A determiner is a grammatical word associated with a noun phrase that helps express how its referent should be understood in context. English examples include the, a, an, this, and that. Determiners can signal distinctions such as definiteness, selection, or quantity. The exact category boundaries depend on the language and grammatical or annotation framework.

What it means

A noun names a category or concept, while a determiner can help frame which instance or set is being discussed. Compare a report, the report, and this report. The central noun stays the same, but the way the speaker presents its reference changes.
An article is one kind of determiner. The English articles a, an, and the do not exhaust the category. Demonstratives such as this and that also occur in determiner uses.
For natural-language processing, the category provides information about sentence structure and reference. It does not by itself identify the exact real-world object. Knowing that someone said the report does not tell a system which report they meant without sufficient context.

How it works

A part-of-speech tagger analyzes a token in its sentence and assigns a grammatical category under a defined tagset. A parser can then represent how the determiner relates to the noun phrase.
The analysis uses grammatical distribution, not just the broad meaning of the word. The same spelling can have different roles in different constructions. For example, that in that folder differs from that in I heard that the folder was missing.
Annotation schemes make explicit choices about difficult boundaries. Universal Dependencies uses DET for articles and certain quantifying or pronominal words, while narrow cardinal numerals receive NUM. Possessive-word classification can vary across languages. A dataset should therefore document its conventions instead of assuming every grammar uses identical labels.

Three practical examples

• A grammar-support tool can use determiner information when examining English noun phrases and possible article errors.
• An information-extraction system can preserve the difference between a newly introduced item and an item presented as identifiable in context.
• A linguistic annotation project can define how determiners, pronouns, and numerals are labeled so that examples are treated consistently.

Think of it this way

Think of a noun as naming a drawer of possible objects and a determiner as a small instruction about how to look inside it: an unspecified member, an identifiable member, or a contextually indicated one. The analogy helps with reference, but determiners are grammatical items with language-specific behavior, not literal database selection commands.

Worked example

Consider the invented sentence: A designer opened the folder and reviewed this sketch. Ignoring punctuation, its words are A, designer, opened, the, folder, and, reviewed, this, sketch.
In a straightforward English analysis, A introduces the noun phrase A designer, the accompanies folder, and this accompanies sketch. These three tokens are determiners. Designer, folder, and sketch are nouns, while opened and reviewed are verbs. And connects the coordinated actions.
Now compare: I heard that the designer opened the folder. Here, that introduces the clause the designer opened the folder. It is not performing the determiner role found in that folder. A system that labels every occurrence of the spelling that as DET would therefore make an error in this example.
Finally compare A designer opened a folder with The designer opened the folder. Both can describe an opening event, but they present the participants differently with respect to identification in context. The definite version does not automatically prove that the two people in a conversation actually share enough information to identify those participants.
This example shows two separate tasks: recognizing grammatical category and resolving reference. Correct determiner tagging helps the first and may inform the second, but does not complete both tasks at once.

Illustrative case study

In a fictional document-assistance project, a preprocessing step removes frequent short words before the system analyzes user requests. The resulting text loses distinctions among open a file, open the file, and open this file.
The team retains the original text for interpretation and uses task-specific preprocessing rather than assuming all frequent function words are disposable. It tests whether references can be resolved from the surrounding conversation and interface context.
The case illustrates why small grammatical words can carry useful information. It does not claim that keeping determiners alone solves ambiguous references.

Common misconceptions and mistakes

Do not equate determiner with adjective. Both can relate to nouns, but their grammatical roles and annotation categories differ.
Do not assume every noun must have an overt determiner. Languages and constructions differ; English also permits bare plurals and mass-noun uses.
Do not classify a token solely from its spelling. Context can distinguish determiner and non-determiner uses.
Do not treat one tagset's treatment of possessives or quantifiers as universal. Also, the presence of the does not mechanically mean the noun was mentioned earlier. Identifiability can arise from shared context or other information, and references can remain ambiguous.

Business use

Determiner-aware analysis can support more faithful language processing in document tools, conversational interfaces, and grammatical feedback. Its value depends on whether the task needs sentence structure or reference distinctions.
When evaluating a parser or tagger, include short function words and ambiguous forms rather than focusing only on prominent content words. Record the annotation scheme and language-specific rules. For user-facing applications, preserve enough context to interpret references instead of expecting a grammatical label to identify an object on its own.

People also ask

Are a, an, and the determiners?

Yes. They are English articles, a type of determiner.

Is that always a determiner?

No. Its role depends on the construction, such as that folder versus I know that it arrived.

Are determiner categories identical in every language?

No. Both language structure and annotation conventions affect the classification.

Check your understanding

In I know that the report arrived, should that automatically receive the same analysis as that in that report?

Answer: No. The first introduces a clause; the second accompanies the noun in a determiner use.

Related terms

• Article (part-of-speech)
• Noun phrase
• Part of speech
• Pronoun
• article
• definiteness
• demonstrative
• reference resolution
• and Universal Dependencies

Learn next

Compare several sentences containing a, the, this, and that. Identify the noun phrases, then separate grammatical tagging from the contextual question of which entity each phrase refers to.

More in Natural language and speech

• Dental sound - A dental sound is a speech sound whose principal constriction involves the tongue at the teeth.
• Dependency tree - A dependency tree represents the syntactic structure of a sentence through directed relations between its words.
• Derivational morphemes - Derivational morphemes are meaningful word-building units that help form new lexical words from a base. They can change meaning, grammatical category or both.
• Development set - A development set is a collection of examples used to guide choices while building a machine-learning system.
• Dialogue act - A dialogue act is the communicative function performed by an utterance or part of an utterance in a conversation.
• Dictionary-based machine translation - Definition: Dictionary-based machine translation translates source expressions by looking up corresponding target-language entries in a bilingual dictionary.

Sources

Sources: Jurafsky and Martin, Speech and Language Processing, Chapter 18, English word classes: https://web.stanford.edu/~jurafsky/slp3/18.pdf Universal Dependencies v2, DET category and language-dependent boundaries: https://universaldependencies.org/u/pos/DET.html

Previous: Determinant Next: Determining the number of clusters in a data set