Word Frequency Counter

What is word frequency, and what is it for?

Word frequency is a count of how many times each distinct word appears in a text, ranked from the most frequent down. A word counter tells you the size of a document; a frequency counter tells you what the document is made of, which is a different question with different uses.

Four uses account for most of the demand. Writers use it to find the word they have leaned on forty times without noticing. Editors use it to check that a term is used consistently rather than in four variants. Students and researchers use it as the first step of any text analysis, because a ranked frequency list is the cheapest description of a corpus that exists. Anyone working on a page uses it to see which words that page actually emphasises, as opposed to the ones they intended to.

How does this tool decide what counts as a word?

A word is a run of letters, digits and internal apostrophes, and everything else is a boundary. That definition sounds obvious and it is where most disagreements between two counters come from, so it is written out here rather than hidden.

• Punctuation splits. Full stops, commas, brackets, dashes and quotation marks all end a word rather than becoming part of one.
• Apostrophes inside a word are kept. don't is one word, not two, and 'quoted' loses its outer apostrophes.
• Hyphenated words split into two. well-written counts as well and written, which is the commonest reason a count here differs from a word processor.
• Accented and non-Latin letters are letters. The tokeniser accepts Latin-1, Latin Extended and Cyrillic ranges, so French, Spanish, Turkish, Polish and Russian text is counted rather than shredded at every accent.

What do the three options change?

Each option answers a question about your text, and the right setting depends on why you are counting. All three are applied before the count, not after it.

Case folding merges The and the into one entry. Leave it on for prose, where capitalisation is a position rather than a meaning. Turn it off when the case carries information, such as code identifiers, product names or a list of proper nouns.

Hiding common words removes about fifty function words such as the, and, of and to. These are always the top of any English frequency list and they say nothing about the subject, so hiding them promotes the words that do. Keep them visible if you are studying style rather than subject, because function-word ratios are exactly what stylometry measures.

Ignoring numbers drops purely numeric tokens. Useful on invoices, tables and anything with dates; harmful when the figures are the point.

How is the share column calculated?

Each word's count is divided by the total number of counted words, after your options are applied. That last clause matters: if you hide common words, the denominator drops, and every remaining word's share rises accordingly.

This is the honest way to report it and it can look surprising. A word appearing twelve times in a 400-word text is 3 per cent of it with function words included, and noticeably more once they are hidden. Neither number is wrong, and they answer different questions, so the tool reports the total it used alongside the table.

As a rough calibration for English prose, no single content word usually exceeds about 2 to 3 per cent of a natural text. A content word well above that is the repetition you came here to find.

Can you get the results out as a file?

Yes. The copy button puts the whole table on your clipboard as CSV, with a header row. Paste it straight into a spreadsheet and the four columns land as rank, word, count and share.

The table shows the top 200 entries, which is where a frequency list stops being readable, while the summary line above it reports the true totals: how many words were counted and how many distinct words exist. Those two numbers together are the useful summary of any text, and the ratio between them is its vocabulary richness.

What do you do with the list once you have it?

A frequency list is a diagnosis, and three findings in it are worth acting on. Reading it top to bottom without knowing what to look for is why most people open one of these tools once and never again.

The first is an unintended favourite: a content word far above the others that you did not choose to emphasise. In a draft it usually means a paragraph is circling; in a published page it means the page is about something slightly different from what you intended.

The second is a term in variants. If e-mail, email and Email all appear as separate rows, the document is inconsistent, and turning case folding off makes that visible rather than hiding it behind a merged count.

The third is a long tail that never repeats. When almost every word appears once, the text is either very short or very unfocused, and the ratio between counted words and distinct words in the summary line is the fastest way to see it.

Is the text uploaded anywhere?

No. Tokenising and counting happen in your browser and nothing is sent to a server. That is the reason this kind of tool is safe to use on a draft, a transcript, a contract or anything else you would not paste into a public form.

If you need the size of a document rather than its composition, our word counter reports words, characters, sentences and reading time. If you need to strip duplicate lines before counting, the duplicate remover does that first.