Word Frequency Counter
Which words you actually overuse — as a ranked table, not a guess.
What the ranked list tells you
A frequency table converts a vague sense of repetition into a number. The two things it reveals most reliably are crutch words — the adjectives and connectives a writer leans on without noticing, usually obvious once ranked — and topic concentration, because a document's subject is whoever appears at the top of the list after stop words are removed.
Stop words change everything, so decide first
With stop words included, the top of every English list is the same: the, and, of, to, a. Those words carry no information about your writing, and leaving them in buries the signal under twenty rows of grammar. Filter them and the picture inverts — suddenly you are looking at what the text is about.
| What you are doing | Stop words | Case | Why |
|---|---|---|---|
| Editing your own prose | Removed | Folded | To surface crutch words |
| Keyword research | Removed | Folded | To find topic terms |
| Checking poetry or lyrics | Kept | Kept | Repetition and case are deliberate |
| Analysing code or logs | Removed | Kept | Identifiers are case-sensitive |
| Language learning | Kept | Folded | Function words are the lesson |
What a simple counter gets wrong
- Morphology.
run,runs,runningandranare four entries. Real text analysis needs stemming or lemmatisation to group them, which is why a word-frequency list is a starting point rather than a linguistic analysis. - Multi-word phrases.
climate changecounts as two words. Finding repeated phrases needs n-gram counting, which is a different question. - Hyphenation and apostrophes. Whether
well-knowncounts as one token or two depends on the splitting rule; this tool splits on whitespace and strips surrounding punctuation, sodon'tbecomesdon'tandwell-knownstays one entry. - CJK text. Chinese, Japanese and Korean do not separate words with spaces, so a whitespace-based counter produces one entry per sentence. Word segmentation requires a dictionary-based tokeniser.
Reading the numbers honestly
Raw counts favour long documents, so compare rates rather than totals when looking at two texts. And be careful with a threshold that looks scientific: a word appearing 1.5 percent of the time is not automatically bad, and a word appearing 0.2 percent of the time is not automatically good. The number is a prompt to look at the text, not a verdict on it.
How to use it
- Paste your text.
- Choose whether to remove common stop words.
- Choose whether to fold case.
- Read the ranked table, and set a minimum count to cut the tail.
Worth knowing
- Splitting is whitespace-based and strips surrounding punctuation.
- Stop-word filtering is what makes the top of the list meaningful.
- Rate, not raw count, is the figure to compare across documents.
- CJK text needs dictionary-based segmentation, not whitespace splitting.
Limitations
- No stemming — inflected forms are counted separately.
- Multi-word phrases are not counted; each word stands alone.
- Stop-word lists are language-specific and imperfect for every language.
- Very long documents are bounded by tab memory.