Home › Language games › Language sources
Language sources, credits and licences
All phrases and wording on the language pages were written by us. We compare them with independent sources so that a phrase is only published when two or more agree. The sources are cross-checks, not copies.
| Source | Licence | How we use it |
|---|---|---|
| Tatoeba (sentence translations) | CC BY 2.0 FR | Used only to cross-check our own phrases. Thanks to the Tatoeba contributors. tatoeba.org. No Tatoeba sentence, audio or sentence ID is republished on this site. |
| Wikivoyage phrasebooks | CC BY-SA 4.0 | Used only to check that our wording matches a published phrasebook. Nothing is copied. |
| Wiktionary | CC BY-SA 4.0 | Used only to check that a word exists in the right language with the right meaning (also for the False Friends game). Nothing is copied. |
| Wikidata | CC0 | Official-language relationships between countries and languages. |
| Unicode CLDR and ICU | Unicode License v3 | Language names in their own language, country names, and the Unicode script property used to check script samples. Copyright © Unicode, Inc. Used under the Unicode License. |
| pinyin-pro | MIT | Hanyu Pinyin romanisation for Mandarin. |
| WanaKana | MIT | Kana to Hepburn romanisation for Japanese. |
| @romanize/korean | MIT | Revised Romanization of Korean. |
| Speech | Your device | Audio, where offered, comes from the speech voices already installed on your own device (the browser Web Speech API). We host no audio. |
How a phrase goes live
Each phrase in each language starts as our own wording. It is published only if at least one independent source (a Wikivoyage phrasebook, a Tatoeba sentence or a Wiktionary entry) agrees with it. The few phrases that only we have written (8 of 504 in total) are held back and sent for native-speaker review instead of being guessed. False friends, script samples and country links go through the same two-source rule, using Wiktionary, the Unicode script property and Wikidata.
Romanisation
Mandarin uses Hanyu Pinyin with tone marks, Japanese uses Hepburn with long vowels marked, Korean uses Revised Romanization, Russian a BGN/PCGN-style table, Ukrainian the 2010 national system and Greek ELOT 743. Languages without a reliable software library (Arabic, Hebrew, Persian, Urdu, Hindi, Bengali, Tamil, Thai) use simplified romanisation we wrote by hand, flagged for native review.
Languages and countries
We do not use flags for languages, because languages are not countries. A language is linked to a country only where it has official status in public data, limited to UN member states, and we leave out places whose status is disputed. Distances in Mystery Phrase measure between the core regions of two languages and are only a guide.
Fonts and audio
The site uses the fonts already on your device, so no font files are served. If a script shows as empty boxes, your device may lack a font for it. Computer voices can mispronounce tones, unwritten vowels and names, so the written script and romanisation are the source of truth.
Review status
No phrase has yet been reviewed by a native speaker. If you speak one of these languages and spot a mistake, please tell us through the contact page. See also privacy and all language games.