Broader language coverage
Our chatbots today understand all the major Indo-European languages. Danish, Swedish, Norwegian, English, German, Dutch, Spanish, French, Italian, Portuguese and Hindi. Not because we have translated all content, but because the underlying language models already know them. The next step is the Uralic languages.
One language family at a time
We don't add individual languages, we open up entire language families. The overview shows what is fully supported, what we are working on right now, and what comes next. Our website is already available in Finnish, but the chatbot's deep language understanding for the Uralic languages is still in active development.
-
Indo-European
Danish · Swedish · Norwegian · English · German · Dutch · Spanish · French · Italian · Portuguese · Hindi
-
Uralic
Finnish · Estonian · Hungarian
-
Global
Chinese · Japanese · Arabic · and more
Why do language families matter?
Languages within the same family share grammatical structure, word order and meaning patterns, and language models learn precisely these patterns. Not just individual words. Indo-European languages therefore generalise well across each other: the structure is familiar.
Cross the family boundary (e.g. from Danish to Finnish) and things change. Uralic languages like Finnish, Estonian and Hungarian use inflectional endings rather than word order to express grammatical relations, and words and sentence structures cannot be translated 1:1.
The risk is that nuances and context are lost along the way if a user asks questions in a language from one family to a chatbot indexed in a language from another. E.g. a Finnish user writing in Finnish to a chatbot indexed in Danish. We solve this by actively adapting the platform's language understanding for each new language family we add.