Often times you might want to target a multi-lingual bot. You can of course use Machine Translation as an integral part of your bot like documented here.
In other cases, you might want to manage the translation and localization for the language understanding content for your bot independently.
Translate command in the @microsoft/bf-lu library takes advantage of the Microsoft text translation API to automatically machine translate .lu files to one or more than 60+ languages supported by the Microsoft text translation cognitive service.
You can learn more about language x locale support for qnamaker.ai here
- An .qna file and optionally translate
- Comments in the qna file
- QnA reference link texts
- List of .qna files under a specific path.
>bf qnamaker:translate
OPTIONS
-f, --force If --out flag is provided with the path to an existing file, overwrites that file
-h, --help qnamaker:translate help
-i, --in=in Source .qna file(s) or QnA maker application JSON model
-o, --out=out Output folder name. If not specified stdout will be used as output
-r, --recurse Indicates if sub-folders need to be considered to find .qna file(s)
--srclang=srclang Source lang code. Auto detect if missing.
--tgtlang=tgtlang (required) Comma separated list of target languages.
--translate_comments When set, machine translate comments found in .qna file
--translate_link_text When set, machine translate link description in .qna file
--translatekey=translatekey (required) Machine translation endpoint key.
qnamaker:translate command expects a Machine translation subscription key. You can obtain one here
You can follow instructions here to create LUIS models from lu files generated via luis:translate command.
Note: You need to explicitly provide the correct QnAMaker lang to the bf qnamaker:translate command.
Note: bf qnamaker:translate command does not verify validity of the .qna file. You might want to try qnamaker:convert with the .qna file(s) before translating to address validity issues in the source language before translating.