The questions are classified according to their language. If questions are english, they're sent to model 2, otherwise they're sent to model 1 (general model).
I evaluated CLD2, CLD3 and LangID for language detection.
CLD2 was by far the best on our questions (accuracy: 98.1%, macro f1: 81.7%, f1 for 'en': 99.06%).
The questions are classified according to their language. If questions are english, they're sent to model 2, otherwise they're sent to model 1 (general model).
I evaluated CLD2, CLD3 and LangID for language detection.
CLD2 was by far the best on our questions (accuracy: 98.1%, macro f1: 81.7%, f1 for 'en': 99.06%).