Is it possible to feed the NLU engine with slot’s information in order to enhance the intent prediction? Let’s illustrate it with an example: the sentence ‘give me the recipe’ could prompt two different answers, an utterance corresponding to a cake recipe and other to a pizza recipe, according to a previously filled slot called ‘food_type’. If the intent classifier could receive this slot along with the sentence (or the tokenized sentence), the intent classification task would be fairly easy to accomplish (get_pizza_recipe or get_cake_recipe), otherwise the user’s intent would be dubious and impossible to predict.
You’re right that using the slot value can help your bot respond more accurately, but the recommended way to do this would be to set slots in your stories to influence action prediction, not intent prediction. So something like this in your stories.md file:
## pizza recipe
* askrecipe
- slot{"food_type":"pizza"}
- utter_pizza_recpe
## cake recipe
* askrecipe
- slot{"food_type":"cake"}
- utter_cake_recpe