I want to know roadmap to build a multilingual chatbot i.e bot could take input as mixture or hindi and english words. Any dataset or refereces to achieve this task would to be highly appreciable.
1 Like
Hello, Are you talking about written Hindi language or the transliteral Hindi as we see on the internet?
There is a difference in both…
If you are doing simply written Hindi, I would recommend a custom language detection component since the characters are very different from English, it works quite easily.
Now let’s talk about transliterals.
You can try the self supervised configuration without pre trained weights and it seems to be working. I tried a loooong time ago for Bengali and it was working but I never extensively tested it. Ofcourse only worked for intent classification for me.
Yeah, I am talking about transliteration. Thanks.
Hi Rajeev, check out this solution on the thread I initiated when I had the same issue.
Basically, there are two approaches described. choose the one that suits you best.
Hmmmm… Thanks