N-gram Extractor
Extract n-gram (word sequence) frequency from text — pick a gram size from 1 (single words) to 5, get a ranked frequency table with counts and percentages, and optionally drop n-grams made up entirely of stop words.
Input
Number of consecutive words per n-gram (2 = bigrams, 3 = trigrams, ...).
Count "Word" and "word" as different words.
Drop n-grams made up entirely of filler words like "of the" or "and the".
Output
| # | N-gram | Count | Percentage |
|---|---|---|---|
| No data yet | |||
Guides
The N-gram Extractor reads a block of text and finds every repeated sequence of consecutive words — bigrams, trigrams and beyond — then ranks them by how often they occur. It's the fastest way to spot recurring phrases, stock language, or the multi-word terms that actually define a text's topic, which a single-word frequency count can't show you.
What it measures
Every n-gram (a sliding window of N consecutive words) is counted and shown in a ranked table with four columns: its rank, the n-gram itself, how many times it appears (count), and what percentage of the counted n-grams it represents. Rows are ordered by count, highest first, with alphabetical ordering as a tiebreaker so the same text always produces the same table.
How to use it
- Paste or type your text into the input box — the table updates live as you type.
- Adjust the options to focus the results:
- N-gram size (N) — 2 finds pairs of words (bigrams), 3 finds triplets (trigrams), and so on up to 5.
- Case sensitive — off by default, so "The" and "the" are treated as the same word inside an n-gram.
- Ignore stop-word-only n-grams — drops sequences made up entirely of filler words like "of the" or "and the", so the ranking surfaces phrases that carry real meaning.
- Show top N results — caps how many rows the table displays.
- Use the copy or download button to export the table as CSV for a spreadsheet or report.
How n-grams are extracted
Text is first split into words using a Unicode-aware rule, so accented and non-Latin scripts are handled correctly, and internal apostrophes or hyphens are kept intact (so don't and well-being each count as one word). N-grams are then built by sliding a window of N words across the token stream, one word at a time, so overlapping sequences are all captured.
Common uses
- SEO and content teams finding the multi-word phrases (long-tail keywords) that dominate a page or article.
- Writers and editors catching repeated phrases or filler expressions.
- Researchers and linguists studying collocations — which words tend to appear together.
- Developers profiling recurring patterns in logs, chat transcripts or generated text.
Is my text private?
Yes. All extraction happens locally in your browser — your text is never uploaded to a server.
Why don't the percentages add up to exactly 100%?
Each percentage is rounded to two decimal places, so a handful of rounded values can total slightly more or less than 100. The underlying counts are always exact.