How to use the sentiment analyzer
- Type or paste a sentence, review, comment or short message into the box.
- Click "Analyze Sentiment". The first analysis downloads the model, which can take a little while; later analyses are fast.
- Read the label — POSITIVE or NEGATIVE — together with a confidence percentage.
- Change the text and analyze again as many times as you like.
How sentiment analysis works
This tool runs a distilled BERT model through Transformers.js, which executes the neural network directly in your browser using WebAssembly and WebGPU where available. The model, distilbert-base-uncased-finetuned-sst-2-english, was trained on movie reviews and learns to judge whether a piece of text expresses a positive or negative feeling. When you analyze text, the model returns a label and a score between zero and one, which is shown here as a percentage.
What the confidence score means
The confidence score is the probability the model assigns to its prediction. A score near 100% means the model is very sure, while a score near 50% means the text was roughly neutral or ambiguous. Short, sarcastic or mixed sentences are the hardest for any sentiment model, so treat borderline scores as uncertain rather than definitive.
When to use it
Support teams use sentiment analysis to spot frustrated customers quickly. Marketers gauge the tone of reviews and social comments. Writers check whether their wording reads as intended, and researchers sort large numbers of short texts by tone. Because it runs in your browser, it is also a safe way to check private or sensitive text you would not want to paste into a third-party service.
Privacy and the first-load delay
The first analysis feels slower because a roughly 70 MB model is downloaded to your device. After that, your browser caches it, so later analyses run almost instantly. Throughout the process your text never leaves your device — it is processed locally and is never stored or transmitted.