Making a word cloud in another language

WordCloud+ is not limited to English. It recognises the language of your text, removes that language's common words, and builds the cloud from what is left. There is one important limit today: it only works for languages written in the Latin alphabet. This article sets out exactly what works, what does not yet, and how to take control of the language yourself.

Languages that work today

You can make a word cloud, with a default stop word list, in these 36 languages: Afrikaans, Basque, Breton, Catalan, Croatian, Czech, Danish, Dutch, English, Esperanto, Estonian, Finnish, French, Galician, German, Hausa, Hungarian, Indonesian, Irish, Italian, Latin, Latvian, Norwegian, Polish, Portuguese, Romanian, Slovak, Slovenian, Somali, Southern Sotho, Spanish, Swahili, Swedish, Turkish, Yoruba and Zulu.

Accented letters are kept as they are, so "café", "crème" and "Straße" appear with their accents. Everything is shown in lower case, as it is for English.

Languages that do not work yet

The Language list in the editor also includes Arabic, Armenian, Bangla, Bulgarian, Chinese, Greek, Hebrew, Hindi, Japanese, Korean, Marathi, Persian, Russian, Thai and Ukrainian. You can select them and their stop word lists will load, but the cloud itself will stay empty. WordCloud+ currently only counts words that contain at least one Latin letter or digit, so text written entirely in Arabic, Cyrillic, Greek, Hebrew, Devanagari, Chinese, Japanese, Korean, Thai or other scripts produces no terms. Chinese, Japanese and Thai are also written without spaces between words, which the word splitting relies on.

A mixed text, such as Russian with some English product names, will produce a cloud of the Latin-alphabet words only.

How the language is chosen

When your text becomes a project, WordCloud+ guesses its language from the text and loads the default stop words for it. If it cannot tell, or the language it finds has no stop word list, it uses English.

The free preview on Create a word cloud always uses English stop words, so a French or German preview may show common words such as "les" or "und". They disappear once you choose Unlock editing: the draft becomes a project and the language is detected then.

Change the language yourself

1. Open the project and select the Settings tab.
2. Open the Language list. Languages are listed by their English names.
3. Choose the language of your text. The cloud is recalculated with the new stop words.

As the setting itself warns, changing the language resets the stop words to the defaults for that language. Any words you added yourself are lost, so if you have customised the list, copy your additions from the Stop Words tab before switching and paste them back afterwards.

English stop words are built into WordCloud+. The lists for other languages come from an open-source stop word collection and are downloaded when needed. If that download fails you will see "Failed to fetch default stop words" and the project is left with an empty stop word list; check your connection and set the language again.

Text in more than one language

A project has one language and one stop word list. For a survey answered in, say, English and Spanish, build a combined list:

1. Set Language to the second language, open the Stop Words tab and copy the whole list.
2. Set Language back to the main language.
3. In the Stop Words tab, paste the copied words below the existing list and select Update.

Stop words must be in lower case and match the text exactly, including accents. How to remove words from a word cloud covers editing the list.

Merging words in other languages

The Replace tab works in any supported language, for example to fold plural and singular forms together. One caveat: it does not currently match a word that begins or ends with an accented letter, such as "café" or "école", though accents in the middle of a word are fine. How to combine and rename words in a word cloud explains how replacements work.

Ready to try your own text? Create a word cloud, then unlock editing so the language is detected.