See more
If possible train Intents with original corpus of conversations, otherwise train with manufactured utterances. Minimum 5 utterances, optimum 10 utterances.