Word Frequency Counter

A word is a run of letters and digits, kept whole across an inner apostrophe or hyphen, so don't, e-mail and follow-up each count once. Counting ignores case: Price and price are the same word. Accented letters stay inside the word, and text is first put in the Unicode composed form (NFC), so a café typed with a separate accent mark is the same word as one typed with é.

The percentage is the word's count divided by all the words in the text. In "The price is fine, but the delivery time worries me. Delivery time is six weeks. The price includes delivery." there are 19 words; delivery appears 3 times, which is 15.79%, and price and time twice each. Stop words and short words you hide still count in the total, so the percentages do not jump when you change a filter.

Numbers are left out by default (switch it off to count them; 3.5 stays one number). The total, the unique words and the number hidden by your filters are shown above the table.

Stop words are the short, frequent words that carry little meaning on their own: the, and, of, is, I, it. Left in, they fill the top of every list. This page hides them using the Snowball stop word list for the page's language (174 English words, including contractions such as don't and it's), a list published by the Snowball project under a BSD license.

Switch the filter off to see them, or add your own words to hide, such as yeah, okay or a name that appears in every line. A list is a starting point, not a rule: in a text about pronouns or in a speech, the stop words may be exactly what you want to count.

The phrase tabs count word pairs and triples, such as delivery time or return on investment. Phrases are counted within a sentence and do not run across a full stop, a question mark, a colon or a line break. With stop words hidden, a phrase that starts or ends with one (of the, the price) is dropped, while one with a stop word in the middle (return on investment) stays.

Only phrases that appear at least twice are listed, since a phrase said once tells you little. In a call or interview transcript, the repeated phrases are often the topics the other person kept coming back to.

Paste an interview, a sales call or a meeting transcript and switch on Ignore speaker labels and timestamps: labels such as Anna: or Speaker 1 and times like 00:03 or WebVTT and SRT cue timings are removed before counting, so speaker names do not top the list. You can also open a .txt, .vtt or .srt file.

A ranked list is a quick way to check what really got talked about: the customer said downtime eleven times and price twice; the interviewee kept saying funding. Search the table for a word you expected to hear and see whether it came up at all.