CanDoYa
EN

Free Sentiment Analysis Tool

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

Paste feedback, a review, or a social post to assess its tone.

Your text is processed in a browser worker and is never uploaded. The first run downloads the public AI model.

Three overall sentiment labels

Negative means unfavorable tone, neutral means no clear positive or negative lean, and positive means favorable tone.

Fine-tuned for Arabic, English, French, German, Hindi, Italian, Portuguese, and Spanish.

Built from social posts. Sarcasm, mixed feelings, specialist language, and text outside its trained languages can be mislabeled.

Share this tool

What is sentiment analysis?

Sentiment analysis classifies the overall attitude in text as negative, neutral, or positive. This tool runs a multilingual XLM-T model inside your browser and shows the score for every label. Your text stays on your device, although the public model downloads when you first run an analysis.

How to use

  1. 1Paste a useful sample. Enter a review, comment, message, or social post. A complete sentence gives the model more context than one word.
  2. 2Run the local model. Click Analyse sentiment. Allow the first model download to finish; later visits can reuse it while the browser cache remains available.
  3. 3Read all three scores. Use the top label as a broad signal and compare its score with the alternatives. Close scores call for human review, not a firm conclusion.

Who it's for

This sentiment analyser uses the Cardiff NLP XLM-T model, an XLM-R system trained on roughly 198 million multilingual tweets and fine-tuned for sentiment in Arabic, English, French, German, Hindi, Italian, Portuguese, and Spanish. The output labels mean unfavorable overall tone, no clear positive or negative lean, and favorable overall tone. They do not identify specific emotions, truth, urgency, or the writer's intent.

The first analysis downloads quantized ONNX model weights plus tokenizer files, about 297 MB in total. A browser worker keeps inference off the main UI thread. WebGPU is preferred when available; the WebAssembly CPU backend is the slower fallback. Browsers normally cache the files, but private browsing, storage cleanup, or low disk space can trigger another download.

FAQ

Is my text uploaded to a server?

No. Your text is sent to a worker inside your browser and the model runs on your device. The worker downloads the public Transformers.js runtime and model files from external hosts, but those requests do not contain your pasted text. CanDoYa does not receive or store the sample.

Is this sentiment analysis tool free?

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

How much text can I analyse?

You can paste an ordinary passage, but the tool analyses at most the first 1,000 Unicode characters to keep browser inference predictable. The model tokenizer also truncates text to its input window. For long documents, analyse coherent sections separately instead of treating one overall label as a document summary.

Which languages does the sentiment model support?

Sentiment fine-tuning covered Arabic, English, French, German, Hindi, Italian, Portuguese, and Spanish. The multilingual base model may return labels for other languages, but those results do not have the same task-specific training evidence. The page therefore marks other CanDoYa locales as approximate.

What do negative, neutral, and positive mean?

Negative means the model found an unfavorable overall tone. Neutral means it found no clear positive or negative lean. Positive means it found a favorable overall tone. These labels describe the whole submitted sample; they are not emotion categories and do not prove what the writer believes or intends.

What do the model scores mean?

The three scores rank negative, neutral, and positive for this input and normally sum to about 100 percent. They are useful for comparing the model's labels, but they are not guaranteed probabilities of correctness. A high score can still be wrong, especially when the text falls outside the training domain.

Why is the first download about 297 MB?

The tool uses quantized XLM-R ONNX weights of about 279 MB and tokenizer data of about 17 MB. That is much larger than a keyword list, but it captures contextual language patterns instead of merely counting positive and negative words. The browser normally caches these files after the first run.

Does the analyser work without WebGPU?

Yes. The worker first uses WebGPU when the browser and graphics adapter support it. If that path is unavailable or fails to start, the tool retries on the WebAssembly CPU backend. CPU analysis uses the same model and keeps text local, but loading and inference can take longer.