I am building a conversational assistant using Rasa and I’m having an issue where Rasa is triggering the wrong intent, even though the intent recognition probabilities indicate a different intent should have been triggered.
- I’m training a Rasa model to handle intents like “hoi_sach” (ask about books), “xem_gio_hang” (view cart), etc.
- When I pass intent probabilities to Rasa, “hoi_sach” has a very high probability of 0.999941349029541
- “xem_gio_hang” has a very low probability of 1.756610072334297e-05
- However, Rasa is triggering “xem_gio_hang” instead of the much more likely “hoi_sach”