Training Data Format, for same intent in multiple languages

In my use case, I have an intent in multiple languages. I’ve been migrating from Dialogflow to Rasa. This is one of my intent, greet data in English greet.json (1.4 KB) I also have the same intent in 7 more languages(20 intents in 7 languages). This is my config file config.yml (189 Bytes) . From the documentation, I understood that tensorflow pipeline works best for multi-language support, but am clueless about data representation for the same. i.e do I add training data of all languages in the same file or do I need to create different bots for different languages as suggested for spacy pipeline over here

1 Like

You should create different bots for different languages, or at least different nlu models. The tensorflow pipeline supports a lot of languages, but does not support multi language

1 Like

How to create different nlu models for different language.Could you explain in detail?

@akelad @fede @Juste How do I send http api request to train for multiple language? I understand for each language, there will be a seperate nlu file. and need to add language in config file. Do I have to hit this HTTP API of model training for all the languages? Wouldnt it create a new model for each time?

How can i make multi language rasa chatbot having at least two languages.

@akelad that means we need to create separate docker for each language that runs on different port?