Skip to content
Kordu Tools Kordu Tools

AI Sentiment Analyzer

AI Runs in browser

Analyze text sentiment with AI. DistilBERT understands context, negation, and sarcasm — runs on-device with no data uploaded.

Last updated 01 Apr 2026

Analyzes text sentiment using DistilBERT — a neural network that understands context, sarcasm, and nuanced language. Returns overall sentiment (positive/negative/mixed) with a confidence percentage plus a sentence-by-sentence breakdown. Runs entirely in your browser after a one-time 67 MB model download.

~63.9 MB download
0 words

Enter text above to analyze its sentiment with AI
Loading rating…

How to use

  1. 1

    Enter or paste your text

    Type or paste the text you want to analyze. Works best with complete sentences — single words may give ambiguous results.

  2. 2

    Click Analyze

    Press the Analyze button to run the AI model. On first use, the 67 MB DistilBERT model downloads and caches — subsequent analyses are instant.

  3. 3

    View the overall sentiment

    See the overall positive, negative, or mixed sentiment result with a confidence percentage and colour-coded confidence bar.

  4. 4

    Review the sentence breakdown

    For multi-sentence text, each sentence is analyzed individually — see exactly where sentiment shifts across your document.

Frequently asked questions

How is this different from the regular Sentiment Analyzer?
The regular analyzer uses the AFINN word list — instant but context-blind. This AI version uses DistilBERT, a neural network trained on millions of text samples. It correctly handles negation ('not bad' → positive), sarcasm, and sentence context that word-list approaches miss entirely.
What is DistilBERT?
DistilBERT is a compressed, distilled version of BERT (Bidirectional Encoder Representations from Transformers), a language model from Google. Fine-tuned on SST-2, it achieves ~91% accuracy on standard sentiment benchmarks while running fast enough for browser-based inference.
Why does it show MIXED results?
DistilBERT outputs positive and negative probability scores. When neither exceeds 65% confidence, the result is labelled MIXED — reflecting genuine ambiguity rather than forcing a binary label. This is common in balanced reviews or nuanced commentary.
Is my text sent to a server?
No. All analysis runs in your browser using WebAssembly or WebGPU. Your text is never uploaded, stored, or shared with any server.
Can it analyze languages other than English?
The DistilBERT model used here is fine-tuned on English text. For other languages, results may be unreliable. Use the standard Sentiment Analyzer or a language-specific model for non-English content.
How accurate is it?
DistilBERT fine-tuned on SST-2 achieves approximately 91% accuracy on the standard sentiment benchmark. It performs well on product reviews, news, and general English prose but may struggle with highly technical, domain-specific, or emoji-heavy text.
Why does it need a 67 MB download?
The DistilBERT model weights are approximately 67 MB. The model downloads once and is stored in your browser's cache — future visits require no re-download unless you clear your cache.
Does it work on mobile?
Yes, on modern mobile browsers that support WebAssembly (Chrome, Safari, Firefox). The initial 67 MB download applies on mobile too — we recommend Wi-Fi for the first load.
What is sentence-level sentiment breakdown?
For multi-sentence inputs, the tool runs DistilBERT on each sentence individually and shows a per-sentence sentiment badge. This lets you identify which specific sentences are positive or negative within a longer review or document.

Traditional sentiment analysis relies on word lists — each word gets a fixed

score, and the results are summed. Fast, but brittle: it misses negation

("not bad"), sarcasm, and context-dependent meaning. AI-powered sentiment

analysis solves these limitations with a transformer neural network that reads

the entire sentence and understands how words relate to each other.

This tool uses DistilBERT, a compact version of Google's BERT model, fine-tuned

on the Stanford Sentiment Treebank (SST-2). It classifies your text as positive,

negative, or mixed with a confidence percentage. The sentence-by-sentence breakdown

helps you pinpoint exactly where positive or negative sentiment originates in longer

passages — invaluable for product feedback analysis, social media monitoring, and

content moderation workflows.

All processing runs entirely in your browser using WebAssembly or WebGPU. After a

one-time 67 MB model download (cached for future visits), analysis is near-instant.

Your text never leaves your device.

Who is this for? Marketing teams analyzing brand sentiment in customer reviews,

product managers scoring NPS survey responses, content writers checking tone before

publishing, and researchers studying opinion mining at scale.

Related tools

Sentiment Analyzer

Analyze text sentiment instantly — see positive, negative, and neutral scores with a per-word breakdown. No AI model required.

Text Similarity Checker

Compare two texts with AI semantic similarity — detects duplicate content and plagiarism with a 0–100% match score.

Language Detector

Detect the language of any text instantly — identifies 187 languages with confidence scores. No AI model required.

Word Counter

Count words, characters, sentences, and paragraphs with reading time, speaking time, and keyword density.

AI Text Summarizer

Summarize articles, essays, and documents with AI. Choose length, get bullet points, see word-count reduction. Runs on-device.