How to combine and rename words in a word cloud

People rarely use one word for one idea. A set of reviews might mention "customer", "customers" and "client"; a British and American team will write both "colour" and "color". A word cloud counts each spelling separately, so one theme is split across several smaller words and none of them looks as important as it is.

The Replace tab fixes this. It holds find-and-replace pairs that WordCloud+ applies to your text before anything is counted, so the variants are added together into a single term. It is available on the free plan once you have an account; the guest preview on Create a word cloud is read-only until you choose Unlock editing.

Merge words step by step

1. Open your project and select the Replace tab.
2. In the first empty row, type the word you want to fold away under Find, for example customers.
3. Under Replace, type the term it should count as, for example customer.
4. Select Add row for each further pair. Several rows can point at the same term: clients, client and customers can all become customer.
5. Select Save. Go back to the Cloud or Summary tab and the merged term now carries the combined count.

To take a pair out again, select Remove beside the row and then Save.

How matching works

• Whole words only. customer in Find does not touch "customers" or "customerservice". Add a separate row for each form you want to merge.
• Case does not matter. "NHS", "Nhs" and "nhs" are all matched. The replacement is shown in lower case, like every term in the cloud, because WordCloud+ lower-cases the whole text before counting.
• Phrases work too. Find can hold several words, so national health service can become nhs.
• Rows run in order, top to bottom. A later row sees the result of the rows above it, so you can chain them, but check that an early row does not change a word a later row is looking for.
• Your text is not changed. Replacements apply to the analysis only. The Text tab still shows exactly what you pasted, and deleting a row brings the original counts back.

Rename a term

A replacement is also the way to change how a word reads in the cloud. Replace nps with net-promoter-score, or an internal code name with the product name your audience knows. Because the cloud is lower case, a replacement cannot add capital letters.

Keep a phrase together

WordCloud+ finds phrases of two to four words automatically, but only when they occur at least twice and fit its phrase rules. To make sure a phrase always appears as one term, replace it with a hyphenated version: Find customer service, Replace customer-service. The hyphenated form is a single word as far as counting goes, so it shows even if it occurs only once, and its occurrences no longer count towards "customer" or "service" on their own. Word clouds with phrases, not just single words explains the automatic phrase detection.

Drop a phrase entirely

Stop words work one word at a time, so they cannot hide a whole phrase without also hiding its words everywhere else. A replacement can: put the phrase in Find and leave Replace empty. That is useful for a question pasted in with every answer, such as "please describe your experience", while "experience" still counts wherever else it appears. For single words, How to remove words from a word cloud is the simpler route.

Limits worth knowing

• Start and end Find with a letter or number. Matching is whole-word, so a Find that begins or ends with a symbol, such as "c++", does not match. Colons, semicolons, underscores, brackets and straight double quotes (") are removed from the text before replacements run, so leave them out of Find.
• Accented letters at the start or end of a word. Find does not currently match a word that begins or ends with an accented letter, such as "café" or "école". Words with accents in the middle, such as "naïve", match normally.
• Replacing with a stop word hides the result. If the Replace term is on the project's stop word list, the merged word will not appear at all.

A sensible order of work

Start with the Summary tab, which lists every term with its score and number of occurrences, and scan it for variants of the same idea. Add replacements for those, then remove the words that carry no meaning with stop words, then style the cloud. Doing the merging first means the sizes you style are the sizes that reflect your data.

Have some text to try it on? Create a word cloud, then open the full editor to start merging.