Keyword Extractor

Stop words are the most common grammatical words in a language: articles (the, a, an), prepositions (in, on, at, by), conjunctions (and, or, but), common verbs (is, are, was, has), and pronouns (I, you, he, she). They carry no meaningful content for keyword analysis. Filtering stop words is almost always the right choice — without it, your keyword list will be dominated by "the" and "is". Disable stop word filtering only if you are doing linguistic analysis where grammatical words matter, such as authorship attribution or style analysis.

Bigrams are two consecutive words analyzed as a single unit. Enable bigrams when your text contains important compound concepts that lose meaning when split into individual words. Examples: "machine learning" (not just "machine" + "learning"), "open source", "neural network", "user experience", "climate change". Bigrams are especially useful for SEO keyword analysis — search queries often contain two-word phrases, and knowing your content contains "machine learning" is more actionable than knowing it contains "machine" and "learning" separately. For short texts, bigrams may produce noise; they work best on 300+ word texts.

Paste your existing article, product description, or page content and extract the top keywords. Compare the results against the keywords you intended to target — if your target keyword does not appear in the top 10, your content probably needs more natural mentions of it. You can also analyze competitor content: paste a competitor's article and see what keywords their writing emphasizes. The bigrams feature is particularly useful here because it surfaces the exact multi-word phrases their text is optimized around. Export the results as CSV to bring into a spreadsheet for further analysis.

No hard limit — the tool handles articles, essays, and long-form content. Very short texts (under 50 words) produce unreliable results because TF-IDF scoring needs sufficient word variety to distinguish important terms from coincidental repeats. For best results, use texts of 200 words or more. At the other extreme, very long texts (100,000+ words) may take a second or two to process but will complete correctly since all processing runs in your browser.

It filters out any keyword shorter than the specified character count. The default (typically 3-4 characters) prevents very short words that slipped through stop word filtering from appearing in results. For technical content, you might lower the minimum to 2 to include abbreviations. For topic discovery, raising it to 5-6 ensures you only see substantive words. Experiment: if your results include two-letter words that are not meaningful, increase the minimum. If you are losing important short technical terms, decrease it.

Yes. Use the Export CSV button to download a file with three columns: keyword, frequency (raw count), and TF-IDF score. This CSV can be opened in Excel, Google Sheets, or any data tool for further analysis. The Copy Keywords button copies just the keyword list (without scores) to your clipboard, suitable for pasting into meta keyword fields, content briefs, or tag inputs.

No. The entire TF-IDF extraction algorithm runs in your browser using JavaScript. Your text, whether it contains proprietary business content, unpublished articles, or client documents, never leaves your device. Nothing is uploaded, logged, or stored.