Datasets:

Tensoic

/

gpt-teacher_kn

like 2

Hey, I understand your concern, but the data is mostly translated using an API and is poor with context sometimes. Will need huge human effort if we want to translate it manually.
Also, it is a cyclic process (until we get a good bard/openai api for generating data), where we improve model by a little -> generate data to train another model -> which further improves the model and so on. (This is hoping it understands some parallels between languages)