Foundations

Parts of Speech

Every word in every sentence belongs to a category, and this app labels each one twice — once coarsely and once in fine detail. Understanding why there are two scales is the key to the first exercise.

What you’ll learn

Name the seventeen coarse part-of-speech categories
Read a fine-grained Penn Treebank tag and say what it encodes
Explain why one word can carry two different-looking labels
Work out a word's category from what it does, not what it looks like

Two labels for every word

The parser gives each word a POS — a coarse, seventeen-value category that works across languages — and a TAG, a fine-grained English-specific label from the Penn Treebank set. The coarse one says what kind of word this is; the fine one adds tense, number, degree and mood.

Take freeze, freezes and froze. All three are VERB on the coarse scale — that is the whole point of a coarse scale. On the fine scale they are VBP, VBZ and VBD, which is where the person and tense information lives.

The same sentence on the coarse scale

A^DETcurious^ADJfox^NOUNwatched^VERBus^PRONquietly^ADV.

…and on the fine scale

A^DTcurious^JJfox^NNwatched^VBDus^PRPquietly^RB.

Working out a word’s category

The single most common mistake is deciding from the word’s shape. English lets the same spelling be several categories — watch is a noun in a nice watch and a verb in they watch. What settles it is the job the word is doing in this sentence.

Try to replace it
If you can swap the word for thing and the sentence still parses, it is a noun. If you can swap it for ran, it is a verb. If for very, an adverb.
Check what it attaches to
An adjective attaches to a noun; an adverb attaches to a verb, an adjective or another adverb. If you cannot find the word it describes, it is probably neither.
Then add the detail for the fine tag
Once you know it is a noun, ask: singular or plural? Common or proper? That gets you from NOUN to NN, NNS, NNP or NNPS. Same for verbs — tense and person choose between the six VB* tags.

The seventeen coarse categories

These colours are the ones the Workbench token table uses, so a category looks the same wherever you meet it in this app.

Every badge in this table and every tag in the list after it is a link. The Word Classes group takes each category in turn — what it is, how the parser decides, and at least one parsed example sentence for every category and every tag, including the ones the parser rarely or never produces.

POS	Name	What it is	Examples
NOUN	Noun	A person, thing, place or idea.	fox, courage, river
PROPN	Proper noun	A name for one particular thing.	Nina, Lisbon
PRON	Pronoun	Stands in for a noun already known.	she, us, everything
VERB	Verb	An action or a state.	watched, freeze
AUX	Auxiliary	A helping verb carrying tense, mood or voice.	has, was, will
ADJ	Adjective	Describes a noun.	curious, heavy
ADV	Adverb	Describes a verb, adjective or other adverb.	quietly, never
DET	Determiner	Introduces a noun and fixes its reference.	a, the, my
ADP	Adposition	A preposition; relates a noun to the rest.	at, through
PART	Particle	A function word with no category of its own.	to, not, 's
SCONJ	Subordinating conj.	Attaches a dependent clause.	because, whether
CCONJ	Coordinating conj.	Joins two equal elements.	and, but, or
NUM	Numeral	A number, written any way.	three, 42
INTJ	Interjection	An exclamation, grammatically detached.	oh, hey
SYM	Symbol	A non-word symbol.	$, %
PUNCT	Punctuation	Never tagged in the exercises.	. , ?
X	Other	Anything that fits nowhere else.	—

The Penn Treebank tags

This is the palette the Tags exercise draws from — every button you will ever be offered is one of these. They are grouped by family, and the colours vary within a family rather than across it, so NN and NNS read as relatives without being the same colour.

You do not need to memorise it. A realistic sentence uses five to nine of these, and the exercise only ever offers you the ones actually present.

Key terms

POS (coarse): The seventeen-value Universal Dependencies category. Language-independent, and what the syntactic diagrams are built from. The kettle whistled. — DET, NOUN, VERB. She left quickly. — PRON, VERB, ADV.
See: Grammar Basics
TAG (fine): The Penn Treebank label — English-specific, and carrying the tense, number and degree information the coarse category throws away. This is what the Tags exercise asks for. The kettle whistled. — DT, NN, VBD: a singular noun and a past-tense verb. Kettles whistle. — NNS, VBP.
Lemma: The dictionary form of a word: watched → watch, foxes → fox. Stored for every word, though no exercise asks for it. The foxes watched us. has the lemmas the, fox, watch, we. She is running late. has she, be, run, late.
Morphology: The grammatical features derived from the fine tag — Tense=Past, Number=Sing. Derived rather than stored independently, which is why correcting a tag corrects the morphology too. She sings. — VBZ gives Tense=Pres and Person=3. They sang. — VBD gives Tense=Past.
See: Modal Auxiliaries

Practice on a real sentence

Tag every word in this sentence. Notice that dogs and cats both take NNS, so that button stays available after the first correct answer — its count tells you how many are left.

“The agency feeds unwanted dogs and cats.”

Parts of speech

Give every word its coarse Universal Dependencies category.

Words 0/7Attempts 0Correct 0Time 0:00

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

The agency feeds unwanted dogs and cats.

Next: how these categories get arranged into a picture. Read Reed-Kellogg Diagrams for the constituency view, then Dependency Parsing for the other one.