I am trying to build a solution where farmers from various states in India would put in their queries in their native languages .
The proposed solution should then be able to match a given query with the queries stored in a knowledge base. Thereafter the closet matching query within the knowledge base along with the corresponding solution should be shared with the farmer.
Now for this to work, we need to have a shared vector space for the variety of queries ( and the languages in which they are posted).
Could I leverage a language agnostic Language Model in ULMFIT? Any guidance would be greatly appreciated.