Low resource and Code-mixed NLP

I am interested in exploring, improving and interpreting methods in the low resource and code-scripting regime of NLP. Recently, I have been trying to understand how strong and robust embeddings in these settings are.

Relevant works here.