Bot Training - Other Languages

Hello All,
Please, I would like to know how is the best approach to train the Bot in another Language.
For example, I have done already the training in Default Language (English) and I have activated another language (Brazilian-Portuguese), following the documentation: https://developer.kore.ai/docs/bots/advanced-topics/building-multi-language-bots/enabling-an-additional-language/#reupload
I have chosen the Option 1: Duplicate the configurations of the default language.
Then I have started to do some training. and I am not sure if it is correct. Please, consider the use case:

Default Language: English
Intent name: Installation

Additional Language: Brazilian Portuguese
Intent name: Installation (cannot be translate)

Going to Testing → Utterance Testing → I type in Portuguese: Instalação (means installation), I have got this results:

Is there a way to match Fundamental Meaning as well? And, how to train NER for other languages?
Thank you very much for your help!

Hi @henry.correia,

Please follow the below steps:

  • Once you enable a language, say Portuguese(Brazillian), you can switch to the language and change the intent names, Train the utterances, patterns, etc specific to this language. These changes made here will not be copied to the default language.

  • Now, change the intent name for the task installation to Instalação.

  • Navigate to NLP >> Patterns and add “Instalação” as a pattern to the intent “Instalação”.

  • Now, go to “Testing” and test the utterance “Instalação”.

You can observe that the Fundamental meaning model is also matched.

Ideally, for a Default language (En), we would be matching through Fundamental meaning model.
In the case of an additional language, the ML model match is considered as the language utterance configuration, patterns, ML training depends on the individual developer, we would be considering the ML model match.

For explicit Fundamental meaning model match for the utterance in the additional language, you will need to add a pattern to the intent in that specific language as shown above.

Kindly go through and let us know if you need any further clarification on the above.

Regards,
Yoga Ramya.

Hello Yoga,
Thank you very much for sending such detailed steps and of course worked perfectly again :slight_smile:


If I understood you correctly:

  1. Default language = ideal intent match is through Fundamental Meaning model
  2. Additional language = ideal intent match would be through ML, patterns, and for explicit FM match, it is needed to add pattern to the intent in the specific language.
    Is that correct?
    Thank you in advance
    Henry

Hi @henry.correia,

Yes, you will need to add Patterns to intents in the additional language for Fundamental meaning model match.

Regards,
Yoga Ramya.

Hi @yogaramya.mendu

Thank you very much!

Regards,
Henry

when i am following your suggestions it is identifying intent while utterence testing but failing while testing on bot window
(upload://a2kN6k4yF0Ee4LxuLnJTL4ipXEw.png)

and the below image gives utterence testing

I have went through https://developer.kore.ai/docs/bots/advanced-topics/multi-lingual/building-multi-language-bots/#Translatable_Bot_Components but could not find any thing related to training for enabling multi-lingul bots

Please let me know what changes need to be done for swedish language

@renisweta1094,

In the chat window, please say “cheat language sv” and test the same utterance.

  • Keep open the chat window debug console while testing and send us the “NL analysis” tab content if you are still facing the issue.

Regards,
Yoga Ramya