Paste text to extract its most important keyphrases with the RAKE algorithm - ranked by score and occurrence count. For semantic SEO, content briefs and topic analysis. Free, no signup.
Copied!
NLP keywords are the words and multi-word phrases that natural language processing identifies as carrying the actual meaning of a text. Modern search engines run this kind of analysis on every page they crawl: instead of matching strings, they extract the entities and keyphrases your writing emphasises and rank you for those. Running the same extraction on your own draft shows you - before Google does - what your page really says.
The tool implements RAKE (Rapid Automatic Keyword Extraction), a well-known NLP algorithm. Your text is split into candidate phrases wherever a stop-word ("the", "of", "and") or punctuation appears. Every remaining word is scored by its frequency and by how often it appears alongside other words (its degree). Each candidate phrase then gets the combined score of its words, multiplied by how many times the phrase occurs in the text - so a phrase that is both content-rich and repeated ranks highest. Everything runs locally in your browser; nothing is uploaded.
An article, a competitor's page, a transcript, or your own draft.
Three words per phrase works for most content. Switch to "repeated phrases only" for long documents.
The top 30 keyphrases appear with their RAKE score and occurrence count - copy them straight into your keyword sheet or brief.
For semantic SEO, compare the extracted phrases against the keyword you want to rank for: if your target phrase and its close variants are not near the top, the page is not actually about what you think it is. For content briefs, run the top-ranking competitor pages through the extractor and hand writers the phrases those pages share. For editing, a keyphrase with a suspiciously high count is usually your repetition problem. The score column tells you how central a phrase is; the count column tells you how often you said it - good copy scores high on the first without inflating the second.
Feed the extractor this sentence: "Flood insurance claims need careful documentation, and a flood insurance claim without photos is usually rejected." With phrase length 3 it returns flood insurance claims and flood insurance claim as the top keyphrases, followed by careful documentation and photos - while "need", "without" and "usually" never appear, because they are connective tissue, not topics. Run the same test on a full article and the top ten rows are a remarkably honest summary of what you actually wrote about.
Words like "the", "with" and "about" appear constantly but describe nothing, so RAKE uses them as natural phrase boundaries rather than counting them. That is why the output contains clean phrases like "flood insurance claim" instead of fragments like "of the claim". The boundary list this tool uses contains about 230 common English stop-words, including light verbs like "make" and "use" that add length to a phrase without adding meaning.