CanDoYa
All tools
Saved toolsYour favorites

Add to favorites

Saved in this browser only.
Private browser utility / Social & Creator

Free Parts of Speech Tagger

English only

Runs entirely in your browser - no upload, no sign-up.

Live workspaceLocal processing

Paste an English sentence or paragraph to label each token by grammatical role.

Up to 3000 characters. Longer text is split into model-safe passages without being uploaded.

Your text is processed in a browser worker and is not uploaded. The first run fetches the public AI runtime and model files.

See grammar in context

Run the English BERT model to label every word and punctuation token, then inspect the annotated passage and detailed table.

The model uses 17 labels, including noun, proper noun, verb, auxiliary, adjective, adverb, pronoun, determiner, adposition, conjunction, particle, number, punctuation, symbol, interjection, and other.

Share this tool
parts of speech tagger / browser utility
01 / Overview

What is a parts of speech tagger?

A parts of speech tagger labels each word or punctuation token by its grammatical role in context, such as noun, verb, adjective, determiner, or pronoun. This tool runs an English BERT model in your browser, presents 17 concise tags, and exports the result without sending your text to an analysis server.

02

How to use

  1. 01
    Enter an English passage

    Paste up to 3,000 characters from a sentence, paragraph, draft, or transcript. Load the example to preview the result format.

  2. 02
    Run the local POS model

    Click Tag parts of speech and allow the first model download to finish. Later visits can normally reuse the browser cache.

  3. 03
    Inspect the context

    Read the colored annotation, compare tag counts, and use the token table to review each label and score.

  4. 04
    Copy or export

    Copy the tab-separated tags or download CSV, then verify ambiguous or important labels before reusing them.

03

Who it's for

  • Students and language learners checking how words function inside a complete English sentence.
  • Teachers and tutors creating token-level examples for grammar lessons and classroom discussion.
  • Writers and editors inspecting patterns such as dense noun phrases, repeated modifiers, or weak verb choice.
  • NLP developers and researchers exporting a quick CSV sample before designing a larger tagging or corpus workflow.

This parts of speech tagger uses an English BERT-base token-classification model converted to ONNX for Transformers.js. It returns 17 standard labels: ADJ, ADP, ADV, AUX, CCONJ, DET, INTJ, NOUN, NUM, PART, PRON, PROPN, PUNCT, SCONJ, SYM, VERB, and X. Hover over an annotated token to see its full label and model score.

The first run downloads a quantized model file of roughly 96 MB with WebGPU or 110 MB with the WebAssembly fallback, plus tokenizer and runtime files. Inference runs in a worker so the page stays responsive. Your text remains in the browser, while the public model files are fetched from external hosts and normally cached for later use.

FAQ

Is my text uploaded to a server?

No. Your passage is sent to a worker inside your browser, not to an analysis API. The first run still needs network access to fetch Transformers.js, tokenizer data, and public model weights from external hosts. Those files are normally cached, and their requests do not contain the text you entered.

Is this parts of speech tagger free?

Yes. There is no account, payment, API key, or tagging quota. Your own device performs inference. The practical costs are the initial model download, local browser storage, memory, and processing time. Your network provider may still charge for the data used to download the model.

How much text can I tag at once?

Each run accepts up to 3,000 Unicode characters. The tool splits longer input into smaller whitespace-safe passages before model inference. For a full document, tag coherent sections separately so sentence context stays intact, then combine the reviewed CSV exports if you need one larger dataset.

Which parts of speech does the model identify?

The model has 17 labels: adjective, adposition, adverb, auxiliary, coordinating conjunction, determiner, interjection, noun, number, particle, pronoun, proper noun, punctuation, subordinating conjunction, symbol, verb, and other. Adposition covers roles commonly called prepositions or postpositions, while X catches tokens outside the main categories.

How accurate is an online parts of speech checker?

Accuracy depends on how closely your text resembles the model's English training examples. Complete, conventional sentences are easier than fragments, unusual names, code, specialist jargon, or creative spelling. Scores rank the model's own predictions but are not calibrated probabilities, so verify labels used for teaching, publishing, or research.

Can the same word receive different POS tags?

Yes. English words often change grammatical role with context. 'Light' can be a noun, adjective, or verb, and 'record' can be a noun or verb. A contextual tagger examines neighboring tokens instead of using a simple dictionary lookup, but genuinely ambiguous sentences can still produce a debatable label.

Does the POS tagger work without WebGPU?

Yes. The worker tries WebGPU first when a compatible graphics adapter is available. If WebGPU is missing or model startup fails, it retries with the WebAssembly CPU backend. Both paths use the same 17-label model, but CPU loading and inference can take longer on low-power devices.

Can I tag languages other than English?

This page is for English text. The model is English-specific and uses an uncased English BERT tokenizer. It may return labels for other languages, but those predictions lack the same training basis and can be misleading. Use a language-specific or multilingual POS model when working with non-English text.