Free tool Word frequency counter for English and Chinese text
Paste survey answers, support chats, reviews or an article and see which words come up most, ranked with counts, share and cumulative share. Leave out filler words, add your own, and export the whole table as CSV for a spreadsheet.
- Words, characters, phrases
- CSV export
- Runs locally
Count word frequency
Paste the text and the table updates as you type. Use the buttons to export the full table or copy it into a spreadsheet.
Any length. One message or answer per line works well for chats and surveys; lines and punctuation also keep phrases from running together.
Phrases are adjacent words in the same sentence; they never cross punctuation or a line break.
Hides rare words from the table and cloud. Totals and the CSV export follow the same filter.
Comma-separated, e.g. brand names or "hi, thanks" from chat greetings. Not case-sensitive.
Total counted
77
Words left after filtering.
Distinct
64
Different words.
Appear once
58
Type–token ratio
83.1%
Distinct ÷ total. Falls as text gets longer.
Notes
- NoteDetected English (145 words).
- Note68 stop word occurrences left out.
- Note58 words appear fewer than 2 times and are hidden. Totals still include them.
Word cloud
Up to 50 of the most frequent words, size by count.
Frequency table
Share = count ÷ total counted. Cumulative share adds up from the top.
| # | Word | Count | Share | Cumulative |
|---|---|---|---|---|
| 1 | refund | 4 | 5.19% | 5.19% |
| 2 | coupon | 4 | 5.19% | 10.39% |
| 3 | order | 3 | 3.90% | 14.29% |
| 4 | Tracking | 3 | 3.90% | 18.18% |
| 5 | exchange | 3 | 3.90% | 22.08% |
| 6 | shipping | 2 | 2.60% | 24.68% |
Version v1.0Updated Update history
What a frequency table tells you
A frequency table is the quickest way to see what a pile of text is about before reading all of it. It shows what is said often, not what matters most, so treat it as a map for where to read.
Count and share
How often each word appears and what share of all counted words it is. Compare shares, not counts, when texts differ in length: 40 mentions in 5,000 words is less than 30 in 2,000.
Cumulative share
Adding shares from the top. If the first 10 words already cover 30%, the text is focused on a few topics; a slow climb means many different topics.
Type–token ratio
Distinct words ÷ total words. It drops as texts get longer, so only compare texts of similar length. Very low values in short text usually mean heavy repetition.
Words that appear once
In support chats and survey answers, the long tail is where unusual problems hide. Lower "Show words that appear" to 1 and scan it.
Frequency is not sentiment. "Refund" appearing often does not say whether refunds are a problem; read the lines that contain it. The browser's find (Ctrl+F) on your source is the fastest way.
Analyse a text in three steps
Works for chat exports, survey open-ended answers, app reviews, interview notes and articles.
-
1
Paste the text
Put one message, answer or review per line. Leave out signatures and greeting templates, or add their words to the ignore list.
-
2
Choose what to count
Words for an overview, 2- or 3-word phrases for topics such as "coupon code" or "return shipping", single characters for exact Chinese character counts.
-
3
Export or copy
Export CSV for the full table, or copy it straight into Excel or Google Sheets to chart it or compare months.
Share = count ÷ total counted words
Total counted is what is left after stop words, ignored words and (by default) numbers are removed, so shares add up to 100%.
- English words are runs of letters or digits;
e-mailanddon'tare one word, a possessive'sis dropped, and case is ignored. The table shows each word in its most common spelling - English stop words: the NLTK English list (198 words, derived from the Snowball project). Chinese: common function characters and words such as 的, 了, 是, 我们, 因为
- Chinese without a dictionary: split at punctuation and function words, find 2–6 character strings that repeat in this text, then match the longest one (unless a shorter one is much more frequent). Leftover runs of up to 3 characters count as one word; longer runs are split into pairs
- Phrases are adjacent words within one sentence. They may contain a stop word in the middle (
return on investment) but not at either end
Share
count ÷ total counted
4 ÷ 77 = 5.19%
Cumulative share
sum of shares down to this row
5.19% + 5.19% = 10.39%
Type–token ratio
distinct ÷ total counted
64 ÷ 77 = 83.1%
Sixteen support chats, counted
The default text is 16 customer messages from an online store. With stop words left out, 77 words are counted, 64 of them different; 58 appear only once.
| Word | Count | Share | Cumulative |
|---|---|---|---|
| refund | 4 | 5.19% | 5.19% |
| coupon | 4 | 5.19% | 10.39% |
| order | 3 | 3.90% | 14.29% |
| Tracking | 3 | 3.90% | 18.18% |
| exchange | 3 | 3.90% | 22.08% |
| shipping | 2 | 2.60% | 24.68% |
Refunds, coupons, order status and exchanges cover a quarter of everything customers wrote: four canned replies would answer most of these chats. "ordered" and "order" are counted separately because words are not reduced to their stem.
How word counts mislead
The numbers are exact for the text you paste. Most mistakes come from what goes into the text and how the table is read.
Leaving templates in
Greetings, signatures and auto-replies repeat in every chat and push their words to the top. Remove them, or add their words to the ignore list.
Comparing raw counts
A longer export always has bigger counts. Compare shares between months or channels, not counts.
Trusting rare Chinese words
Without a dictionary, words that appear once can be split wrongly. Read the frequent words with confidence and check rare ones against the text.
Reading frequency as importance
A complaint written once by a large customer can matter more than ten mentions of shipping. Use the table to decide what to read, not what to conclude.
Where the text comes from
The most useful text to count is what your visitors and customers write themselves. TapCub collects two kinds of it next to your analytics.
Chat conversations
Export a conversation as JSON or CSV from TapCub live chat, then paste the messages here to find the questions that come up most.
Learn moreSearch queries
Google Search Console and Bing Webmaster queries per page: the words people use to find you, next to the words you wrote.
Learn moreSite search
What visitors type into your own search box, reported with your other page data.
Learn moreWord frequency questions
Still have a question?
Chat with the team behind TapCub — we usually reply within a few hours on business days.
What is a word frequency counter?
A tool that lists every word in a text with how many times it appears, sorted from most to least frequent. This one also shows each word's share of the total, a cumulative share, and a word cloud, and exports the full table as CSV.
How is this different from a keyword density checker?
The keyword density checker starts from the keywords you want a page to rank for and shows their density and positions. This tool has no target: it ranks every word, with stop word controls and export, for analysing surveys, chats or any large text.
What are stop words, and can I add my own?
Stop words are very common words such as "the", "is" and "and" that appear in every text whatever its topic. They are left out by default using the NLTK English list. Add your own in "Also ignore these words", for example your brand name or greeting phrases.
Can it count phrases?
Yes. Choose 2-word or 3-word phrases. Phrases are counted within a sentence and never across punctuation, and a phrase cannot start or end with a stop word unless you include stop words.
Does it work for Chinese?
Yes. Chinese is split without a dictionary, using punctuation, function characters and strings that repeat in the text. Frequent words come out reliably; for exact counts choose "Single Chinese characters".
Can I open the result in Excel?
Yes. "Export CSV" downloads every row with a BOM so Excel reads non-English characters correctly; "Copy table" copies tab-separated rows you can paste straight into Excel or Google Sheets.
Why are "order" and "ordered" counted separately?
Words are counted as written; they are not reduced to a common stem. Stemming merges words that should stay apart, such as "news" and "new". If you need them together, add their counts in the spreadsheet.
Is my text uploaded?
No. Counting runs in your browser. The text is not saved, not added to the page address and not sent to TapCub.
Update history
New features, fixes and data refreshes for this tool, newest first.
-
· v1.0
LaunchedFirst release.
Missing something or found an error? Tell us — requests that come up often go into the next update. All tool updates
Hear what customers ask, as they ask it
TapCub live chat runs from the same snippet as analytics, so every conversation opens with the page, source and path of the visit. Answer there, and export conversations when you want to count them.