What is a part-of-speech tagger?

A part-of-speech (POS) tagger reads a sentence and labels every word with its grammatical category — noun, verb, adjective, adverb, pronoun, preposition and so on. This tool uses the 17-tag Universal Dependencies set, colour-codes the words, and shows a count of each tag so you can see a sentence's structure at a glance.

How do I find the parts of speech in a sentence?

Type or paste the sentence into the box above and tap “Tag text”. Each word is labelled and colour-coded instantly, with a per-word table giving the tag and its definition. You can turn on fine-grained tags for the Penn Treebank codes (NN, VBD, JJ) or download the whole breakdown as a CSV.

What are the 8 parts of speech in English?

School grammar teaches eight: noun, pronoun, verb, adjective, adverb, preposition, conjunction and interjection. This tool uses a finer 17-tag system that splits some of those — proper nouns, auxiliaries, determiners, numerals, coordinating vs subordinating conjunctions and particles each get their own tag — which is the standard used in modern computational linguistics.

What is the difference between POS tagging and named entity recognition?

POS tagging labels the grammatical role of every word — “Colombo” is a proper noun, “is” an auxiliary. Named entity recognition (NER) instead pulls out and classifies real-world names — “Colombo” as a location, “Google” as an organisation — and ignores ordinary words. They answer different questions; we have a separate NER tool for the second.

Is there a free online part-of-speech tagger with no login?

Yes — this one. There is no sign-up, no paywall and no API key. The tagger runs entirely in your browser using an open-source engine, so the text you paste never leaves your device and is never stored. You can use it as often as you like.

Does my text get uploaded anywhere?

No. The tagging model is downloaded to your browser the first time you tag text, and every sentence after that is processed on your own device. Nothing you type is sent to a server, logged or stored. That makes it safe for homework, drafts and anything you would rather keep private.

Which tagset does this use — UPOS or Penn Treebank?

The default view uses the Universal Dependencies UPOS tagset — 17 tags such as NOUN, VERB, ADJ and PROPN. Turn on “fine-grained tags” to also see the Penn Treebank code (NN, NNS, VBD, JJR). The Penn tag is derived from the word's form and is a best-effort approximation; the UPOS tag is always the authoritative one.

Why did it tag a word differently from what I expected?

English is ambiguous — “book” can be a noun or a verb, and the tagger decides from context. It is accurate on ordinary sentences but not perfect, especially on fragments, headlines or unusual phrasing. If a tag looks wrong, add more of the surrounding sentence so the engine has context, or email me the example.

Does it work for languages other than English?

This version is English-only. The bundled engine is trained on English, so tags for other languages would be unreliable. Sinhala, Tamil and other multilingual tagging would need a much larger model, which would slow the page down, so it is kept out of this version.

AI · Language

Part of Speech Tagger

Paste any English sentence and label every word — noun, verb, adjective, adverb and more. The tagger colour-codes your text, counts each part of speech, and runs entirely in your browser, so nothing you type is uploaded. Free, no signup, sources cited below.

By Induwara Ashinsana— Executive Director, Ryzera TechnologiesUpdated Jul 11, 2026

Tag every word17 Universal POS tags

Runs in your browser

Enter a sentence or paragraph

English only. Tagging is deterministic — same input, same tags.44 / 10,000

Try a sample

Toggles above re-render instantly once text is tagged — no re-run needed.

What this does

Paste any English sentence and it labels every word with its part of speech — noun, verb, adjective, adverb, and eleven more — then colour-codes the text, tallies each tag, and lets you copy or download the result. The tagger runs entirely in your browser, so nothing you type is uploaded.

Method: deterministic lexical tagging with wink-nlpfollowing the Universal Dependencies UPOS tagset. Fine-grained Penn tags are derived from word form and are a best-effort approximation. Sources linked under “Sources” below.

How it works

A part-of-speech tagger assigns each word a grammatical category. This tool is deterministic lexical classification — not a probabilistic API call — so the same sentence always produces the same tags, and every example on this page reproduces exactly against the engine you are using.

Tokenise.The text is split into tokens — words, numbers and punctuation — using standard English word-boundary rules. Contractions split the way linguists expect (“don't” → “do” + “n't”) and hyphenated compounds stay together.
Tag. Each token is assigned one of the 17 Universal POS (UPOS) tags by a lexicon-and-rule engine bundled in your browser. There is no temperature and no sampling, which is why the output is reproducible and can be hand-verified.
Label. Each UPOS code is mapped to a friendly name and a definition taken from the Universal Dependencies tagset — ADP → “Adposition (preposition/postposition)”, PROPN → “Proper noun”.
Optional fine-grained view. Turning on fine-grained tags adds the Penn Treebank code (NN/NNS, VB/VBD/VBG/VBZ, JJ/JJR/JJS). Because the engine emits UPOS only, the Penn tag is derived from the word's form — plural nouns from the lemma, verb tense from the ending — and is a best-effort approximation rather than a second trained model.
Aggregate.The distribution counts each tag in a single pass. Each tag's share is count / total over the counted tokens, rounded to one decimal. With “group punctuation” on, punctuation and symbols are left out of both the table and the percentage base.

As an independent check, the tool re-tags a curated list of unambiguous function words — “the”, “and”, “of”, personal pronouns — from a small reference lexicon and reports how many matched the engine. This is a second method that should always agree with the first, the same idea as a calculator cross-checking a total two different ways. Everything runs client-side over WebAssembly and JavaScript; no text is transmitted.

Worked examples

“She quickly wrote three long emails before lunch.”

ShePRONquicklyADVwroteVERBthreeNUMlongADJemailsNOUNbeforeADPlunchNOUN.PUNCT

Distribution: NOUN ×2 (25.0%), then PRON, ADV, VERB, NUM, ADJ and ADP one each (12.5% each), over an 8-word base.

The two nouns are “emails” and “lunch”. “three” is a cardinal number (NUM), not an adjective, and “before” works as a preposition (ADP) here, not a conjunction. The period is punctuation, excluded from the percentage base.

“Colombo is the largest city in Sri Lanka.”

ColomboPROPNisAUXtheDETlargestADJcityNOUNinADPSriPROPNLankaPROPN.PUNCT

Distribution: PROPN ×3 (37.5%) — the three names — then AUX, DET, ADJ, NOUN and ADP one each (12.5%), over an 8-word base.

“Colombo”, “Sri” and “Lanka” are proper nouns (PROPN), each a specific name. “is” is a copular auxiliary (AUX), and “largest” is a superlative adjective (ADJ) — with fine-grained tags on it reads JJS.

“Wow, they have finished the difficult project already!”

WowINTJ,PUNCTtheyPRONhaveAUXfinishedVERBtheDETdifficultADJprojectNOUNalreadyADV!PUNCT

Distribution: INTJ, PRON, AUX, VERB, DET, ADJ, NOUN and ADV one each (12.5%), over an 8-word base. The two punctuation tokens are excluded.

“Wow” is an interjection (INTJ) — a complete utterance on its own. “have” is the auxiliary and “finished” the main verb of the present-perfect “have finished”. With grouping on, the comma and exclamation mark drop out of both the table and the percentages.

Frequently asked questions

Sources & references

The tagset labels, definitions and engine behaviour on this page were last cross-checked against these sources on 2026-07-11. The two worked examples reproduce exactly against the shipped engine.

Related tools

LiveAI

AI Similarity Checker

Compare two texts and get a 0–100% semantic similarity score from sentence-transformer embeddings — catches paraphrases that word-overlap checkers miss. Server-side inference, lexical Jaccard cross-check, no signup, sources cited.

Open tool

LiveAI

AI Resume ATS Checker

Paste a resume and a job description to see an ATS score, missing keywords, format issues parsers struggle with, and section gaps. Scoring follows Workday, Greenhouse, and Lever guidance plus Jobscan's 2024 threshold. Runs in your browser, no signup.

Open tool

LiveAI

Paraphrasing Tool

Reword English text in four modes with a seq2seq paraphrase model, plus a deterministic word-overlap and change-metrics cross-check — free, no signup, sources cited.

Open tool

Rate this tool

Be the first to rate

Comments & feedback

Spotted a bug or want an improvement? Tell us — our team reviews every comment, and good ideas get built. Comments are public and anonymous.

Found a mis-tag, edge case, or want to suggest an improvement?

Email me at [email protected] — most fixes ship within 24 hours.