Word Classes

Determiners & Function Words

The words with little meaning of their own and a great deal of grammar. Five coarse categories — determiners, prepositions, the two kinds of conjunction, and particles — share eight fine tags, and the sharing is where the difficulty lives: that can take three different tags, to two, and IN belongs to two categories at once.

What you’ll learn

Tell a determiner from a pronoun, and a predeterminer from a determiner
Read IN as either a preposition or a subordinating conjunction from what follows it
Split to, off and 's into the right particle tags
Give that its correct tag in each of its three jobs

Closed classes

Nouns and verbs are open classes — English invents new ones every year. The categories on this page are closed: the determiners, prepositions, conjunctions and particles are fixed lists, and you already know every member. So the parser is not really guessing what the word is; it is deciding which of a word’s known jobs it is doing in this sentence.

That is why the same spellings keep coming up. that is a determiner in that pan, a relative word in the tools that we borrowed, and a subordinator in I doubt that the shop is open. Three tags for one word, and the only way to choose is to look at what comes next.

Five categories, eight tags

DETDeterminer

A word that introduces a noun and fixes its reference — which one, how many, whose. The articles a, an, the; the demonstratives this, that, these, those; quantifiers like every, some, no, each; and the question word which when it precedes a noun. A determiner cannot stand alone: every needs a noun after it.

How spaCy decides: a closed list, plus what follows. A possessive like her or their is not DET here but PRON — see PRP$ — while whose goes the other way and is DET. The line is drawn by the model, not by a grammar book, and the drills follow the model.

Every^DETguest^NOUNbrought^VERBa^DETdish^NOUNfor^ADPthe^DETtable^NOUN.

Which^DETcoat^NOUNis^AUXyours^PRON?

DTDeterminer

The ordinary determiner: articles, demonstratives and most quantifiers. It is the third most common tag in the corpus, because nearly every noun phrase begins with one.

How spaCy decides: the word is on the list and a noun follows. When a demonstrative stands alone (That was the worst storm) it keeps the tag DT but its coarse category becomes PRON — the fine scale records the word, the coarse one the job.

The^DTdog^NNchased^VBDa^DTsquirrel^NNaround^INthis^DTyard^NN.

Some^DTbirds^NNSfly^VBPsouth^RBeach^DTwinter^NN.

No^DTbus^NNstops^VBZon^INthis^DTroad^NN.

WDTWh-determiner

A wh-determiner: which and that when they open a relative clause (the book which you lent me, the tools that we borrowed), and which before a noun in a question or an indirect question (which road leads to the coast).

How spaCy decides: which or that immediately after a noun and before a clause. On the coarse scale the relative use is PRON (it stands for the noun inside its clause) and the before-a-noun use is DET; the corpus has one of each.

The^DTbook^NNwhich^WDTyou^PRPlent^VBDme^PRPwas^VBDsuperb^JJ.

The^DTtools^NNSthat^WDTwe^PRPborrowed^VBDbelong^VBPto^INSam^NNP.

PDTPredeterminer

A predeterminer — a word that sits before a determiner and scales the whole phrase: all the cookies, half the class, such a day, quite a crowd, both the pies. The test is the position: the same all directly before a noun (all cookies) is plain DT.

How spaCy decides: one of that small set followed by a, the or a demonstrative. Not before a possessive — both my parents comes back DT. There is none in the corpus yet.

All^PDTthe^DTcookies^NNSvanished^VBDbefore^INnoon^NN.

Half^PDTthe^DTclass^NNwas^VBDabsent^JJon^INMonday^NNP.

Such^PDTa^DTday^NNdeserves^VBZa^DTcelebration^NN.

ADPAdposition (preposition)

An adposition — in English, a preposition: a word that relates a noun phrase to the rest of the sentence, giving place (under the porch), time (during the storm, until noon), direction or possession. The noun phrase after it is its object, and together they form a prepositional phrase.

How spaCy decides: a word from the preposition list with a noun phrase after it. Its fine tag is nearly always IN. The one other tag that maps to ADP is RP, the particle of a phrasal verb — took off — which the coarse scale files as a preposition even though it has no object. See Prepositional Phrases.

The^DETcat^NOUNslept^VERBunder^ADPthe^DETporch^NOUNduring^ADPthe^DETstorm^NOUN.

Two prepositions, one of place and one of time:

She^PRONwalked^VERBthrough^ADPthe^DETpark^NOUNafter^ADPlunch^NOUN.

INPreposition or subordinating conjunction

Preposition or subordinating conjunction — the Penn tag set gives both one tag, which is why its name is a pair. The reasoning is that the two words do the same thing at different sizes: a preposition introduces a noun phrase, a subordinator introduces a clause. before noon and before we left use the same before.

How spaCy decides: the tag is a lookup; the interesting decision is the coarse category, and there the parser looks past the word. Followed by a noun phrase it is ADP; followed by a clause it is SCONJ. The chips below show the preposition reading; the clause reading is under SCONJ just below.

The^DTkeys^NNSare^VBPunder^INthe^DTmat^NNnear^INthe^DTdoor^NN.

The^DTmeeting^NNlasted^VBDuntil^INnoon^NN.

SCONJSubordinating conj.

A subordinating conjunction — a word that attaches a dependent clause to the clause it depends on: because, if, although, whether, that, and the wh-adverbs when, where, why, how. The clause it introduces cannot stand alone as a sentence.

How spaCy decides: a conjunction-like word followed by a subject and a verb. Its fine tag is IN for the ordinary subordinators and WRB for the wh-adverbs — this model puts when and where in SCONJ rather than ADV in every corpus instance. In the dependency tree it carries the label mark; in a Reed-Kellogg diagram it sits on the dashed line joining the two clauses.

We^PRONstayed^VERBinside^ADVbecause^SCONJthe^DETwind^NOUNwas^AUXfierce^ADJ.

The^DETice^NOUNwill^AUXmelt^VERBif^SCONJthe^DETsun^NOUNcomes^VERBout^ADP.

Call^VERBme^PRONwhen^SCONJthe^DETparcel^NOUNarrives^VERB.

The same tag, IN, on its subordinating side — a clause follows in each case, and the coarse category changes to match:

We^PRPleft^VBDearly^RBbecause^INthe^DTroads^NNSwere^VBDicy^JJ.

Although^INit^PRPrained^VBD,the^DTpicnic^NNwent^VBDahead^RB.

I^PRPdoubt^VBPthat^INthe^DTshop^NNis^VBZopen^JJ.

CCONJCoordinating conj.

A coordinating conjunction joins two things of equal rank — two nouns, two verbs, two adjectives, or two whole clauses. The FANBOYS: for, and, nor, but, or, yet, so. If you can swap the two halves and the sentence still works, the joint is coordination.

How spaCy decides: one of the seven, between two like elements. The leading half of a correlative pair — neither in neither the cat nor the dog, both in both knows and corresponds — is also CCONJ/CC, which is how the corpus tags it. Its dependency label is cc, or preconj for that leading half — see Coordination.

Rain^NOUNfell^VERB,yet^CCONJthe^DETcrowd^NOUNstayed^VERBand^CCONJcheered^VERB.

The conjunction joining two whole clauses, with the comma before it:

We^PRONpacked^VERBthe^DETcar^NOUN,and^CCONJthe^DETdog^NOUNjumped^VERBin^ADP.

CCCoordinating conjunction

The coordinating conjunction tag. The only tag in the CCONJ family, and the only fine tag CCONJ ever takes, so the two scales agree completely here — one of the few places they do.

How spaCy decides: the word list. The ampersand is also CC, and so — surprisingly — is a bare plus sign between two nouns (flour + water); the parser reads + as and.

Sam^NNPbought^VBDapples^NNSand^CCpears^NNSbut^CCno^DTplums^NNS.

Neither^CCthe^DTcat^NNnor^CCthe^DTdog^NNstirred^VBD.

PARTParticle

A particle — a function word that fits no other category: the infinitive marker to, the negator not (and n’t), and the possessive ending ’s. They have no meaning to describe, only a grammatical effect, and each is a single fixed word.

How spaCy decides: the three cases are each a lookup. What is worth knowing is the fine tags underneath: to is TO, ’s is POS, and not is RB — an adverb tag on a particle, the one member of this family whose fine tag lives on another page. The phrasal-verb particle off in took off is RP but is not PART on the coarse scale; it is ADP.

Nora^PROPNdecided^VERBnot^PARTto^PARTanswer^VERBthe^DETphone^NOUN.

Sara^PROPN's^PARTbrother^NOUNdoes^AUXnot^PARTlike^VERBolives^NOUN.

TOInfinitive marker 'to'

The infinitive marker to — the to of to travel, to relax, to answer. It has its own tag because the Penn set separates it from the preposition to of to Peru, which is IN.

How spaCy decides: to followed by a base-form verb is TO; to followed by a noun phrase is IN. The first example has one of each. See Infinitives.

Liam^NNPhopes^VBZto^TOtravel^VBto^INPeru^NNP.

The^DTdentist^NNasked^VBDme^PRPto^TOrelax^VB.

RPParticle

The particle of a phrasal verb — the off of took off and turn off, the up of gave up. It looks like a preposition but has no object of its own; it belongs to the verb, and turn off the light can be reordered to turn the light off, which no real preposition allows.

How spaCy decides: a preposition-shaped word directly after a verb that it recognises as forming a phrasal verb with it. Coarse category ADP, not PART, in every instance measured — the tag says particle, the category says preposition. Its dependency label is prt.

The^DTplane^NNtook^VBDoff^RPbefore^INsunrise^NN.

Please^UHturn^VBoff^RPthe^DTlight^NN.

The^DTengine^NNgave^VBDup^RPhalfway^RBhome^RB.

POSPossessive ending

The possessive ending ’s — and the bare apostrophe of a plural possessive like the wolves’ den. The parser splits it from its noun, so mayor’s is two tokens: mayor (NN) and ’s (POS). On screen the two are drawn joined, but in the Tags exercise you click each half.

How spaCy decides: an ’s after a noun that is not a contracted is or has. That’s her car has an ’s that is a verb (VBZ); Jay’s car has one that is POS. Its dependency label is case, and the diagram fuses it back onto its noun. See Possessives & Appositives.

The^DTmayor^NN's^POSspeech^NNran^VBDlong^RB.

The^DTchildren^NNS's^POSlaughter^NNfilled^VBDthe^DThall^NN.

James^NNP's^POSbicycle^NNhas^VBZa^DTflat^JJtyre^NN.

Key terms

Closed class: A word class with a fixed membership — determiners, prepositions, conjunctions, particles. The parser looks the word up and then decides which of its jobs it is doing. The kettle sat on the stove. — the and on belong to closed classes. She left because it rained. — so does because.
Determiner: A word that introduces a noun phrase and fixes its reference. DT, WDT or PDT. The kettle whistled. Which road floods? Both my parents sing.
See: Articles, Adjectives & Objects
Preposition: A word relating a noun phrase to the rest of the sentence. Coarse ADP, fine IN. She sat on the chair. A letter from my aunt arrived.
See: Prepositional Phrases
Subordinating conjunction: A word that makes a clause dependent on another. Coarse SCONJ, fine IN or WRB. We left because it rained. Ask whether the shop is open.
See: Adverb Clauses
Phrasal verb: A verb plus a particle acting as one unit — take off, give up. The particle is RP. The plane took off late. She looked the word up.
See: Function Word Relations

Practice on a real sentence

Two prepositional phrases, so IN and DT each appear twice — the button stays live after the first correct answer and its count tells you how many remain. Both prepositions are ADP on the coarse scale; a noun phrase follows each.

“She sat on the chair in the corner.”

Parts of speech

Give every word its coarse Universal Dependencies category.

Words 0/8Attempts 0Correct 0Time 0:00

Sentence — select a part of speech below, then click the word(s) it applies to

She sat on the chair in the corner.