I would like to use one setup language, for all possible multilingual user inputs.
“Universal Language” is our NLU language where multilingual understanding is fully supported.
Some of the underlying language models we use are indeed, to some extent, multilingual. For example, for German, Arabic or Japanese we use the same model for the system slot parsing, thus it is possible to get the same NLU outcome across these languages and many other of our frontend NLU languages.
This is, however, part of the product architecture and may change over time together with the product improvements. We recommend separating the models for different languages into locales and localizing to ensure the accuracy of recognition. It is even easier to do now, with Real-time translation and Auto-translation features available in the product.
Read more about NLU language support here: https://docs.cognigy.com/ai/nlu/language-support/#overview-of-supported-languages