I'd like to improve the language detection on lesser known languages

Hello David,

Thank you for adding this package, extremely helpful! I'm not a C++ developer, so I was wondering if you could give some insights (or update the readme to give some insights) on how to train the model / improve it? It is lacking quite a bit in accuracy for some of the lesser known languages and want to try to improve it.

Sincerely,
Kevin