About Bigram Frequency
Bigram Frequency counts how often each pair of letters appears in a text and lists the most common pairs one per line - a quick way to see the letter pair profile of a passage. The ranking updates as you type.
Before counting, the text is lowercased and every character outside a through z is removed, so digits, punctuation and spaces all disappear. A two character window then slides one letter at a time, so neighbouring pairs overlap: a six letter word gives five windows, though a recurring pair is counted once. Separators go first, so pairs form across word boundaries too - hello world yields ow and wo from the gap between its two words.
Two limits matter. Only unaccented Latin letters survive, so an accented letter is deleted rather than folded, and a non-Latin script returns nothing. The list stops at twenty pairs without reporting how many distinct pairs existed, so a long text is truncated silently.
How to use Bigram Frequency
- 1 Type or paste the text whose letter pairs you want to count.
- 2 Read the ranking as it updates - the most frequent pair first, one pair per line.
- 3 Copy what you need; remember the list is capped at the twenty most frequent pairs.
Frequently asked questions
What exactly is counted as a pair?
Two consecutive letters after the text has been lowercased and stripped of everything outside a through z. The window slides one character at a time, so neighbouring pairs overlap and a word of six letters produces five windows - though a pair that recurs is counted once, so the number of lines can be lower than the number of windows.
Why do pairs appear that span two words?
Spaces and line breaks are removed before the window slides, so a pair can bridge the gap between two words. In hello world the pairs ow and wo both come from the space in the middle.
Why does my accented or non-Latin text return nothing?
Only the twenty-six unaccented Latin letters are kept. An accented letter is deleted instead of being folded onto its base letter, and text written in Cyrillic, Greek, Arabic, Chinese or another non-Latin script has nothing left to count, so the result is empty.
How many pairs are shown, and in what order?
At most twenty, ordered by count descending. Pairs with the same count keep the order in which they first appeared rather than being sorted alphabetically, and the output does not say how many distinct pairs were found in total.
Free, online, and private
Free to use
Bigram Frequency is completely free. There is no account to create, no login, and no hidden charge or usage limit.
Runs online in your browser
It is a free online tool that runs entirely in your browser. Nothing to download or install — open the page and start using it right away.
Your data stays private
All processing happens locally on your device. Your input is never uploaded, stored, or tracked, so it never leaves your browser.
How Bigram Frequency behaves in detail
These are the specifics of this implementation — the exact ranges, defaults and fixed settings you are working with.
| Parameter | Value |
|---|---|
| Input | one text panel, recalculated live on every keystroke; the sample reads hello world |
| Cleaning | the text is lowercased and every character outside a through z is removed before counting, so digits, punctuation, spaces and line breaks all disappear |
| Window | a sliding window two characters wide moves one character at a time, so neighbouring pairs overlap - a six letter word gives five overlapping windows, though a pair that recurs is still counted once |
| Word breaks | because separators are removed before the window slides, pairs are also formed across word boundaries - hello world yields ow and wo from the gap between the two words |
| Ordering | pairs are sorted by count descending, and pairs that tie keep the order in which they first appeared rather than being sorted alphabetically |
| Output | one pair per line as two letters, a colon and the count, capped at the twenty most frequent pairs; the sample yields nine lines, every one of them with a count of one |
Compiled from the implementation of this tool and checked against it by itoolnet.
What Bigram Frequency does not do
Knowing the edges of a tool saves more time than another list of its features. These are the cases where it will not do what you might expect.
- Only the twenty-six unaccented Latin letters survive the cleaning step. An accented letter is deleted rather than folded onto its base letter, and text written in Cyrillic, Greek, Arabic, Chinese or any other non-Latin script produces an empty result.
- The list stops at twenty pairs and the output never reports how many distinct pairs were found, so on a long text the ranking is truncated with nothing to indicate that anything was left out.
- Only raw counts are reported. There are no percentages, no total number of pairs, no way to keep pairs from spanning word boundaries, and no control over the pair length - longer runs such as trigrams cannot be produced here.
Tested and documented by itoolnet, the team that builds and maintains this site.