Entity mapped to a value problem

Hi, Currently want to check on how can I map a entity which is extracted and map it to a value? For example,

I have different types of pizzas which I have identified them as an intent and categorised them under pizza as shown in nlu.yml :

-intent: pizza_type

examples: |

  • - [cheese pizza](pizza)
  • - [chicken pizza](pizza)
  • - [seafood pizza](pizza)
  • - [super supreme pizza](pizza)

In the actions.py file I map each pizza to a price using a dict as shown:

pizza_prices= { “cheese pizza”:"$12 per pizza", “chicken pizza”: “$15”, “seafood pizza”: “$16”, “super supreme pizza”: “$18” }

class ActionShowPrice(Action):

def name(self) -> Text:

    return "action_show_price"

def run(self, dispatcher: CollectingDispatcher,
        tracker: Tracker,
        domain: Dict[Text, Any]) -> List[Dict[Text, Any]]:
    pizza= tracker.get_slot("pizza")
    price= pizza_price.get(pizza.lower())

    if price is None:

        output = "Sorry could not find"

        output=f"The detail are as follows:\n -Price: {pizza.upper()}\n Price: {price}"

    return []

So my issue now is :

  • If user inputs “cheese pizza” the entity is correctly identified and the output “$12 per pizza” is shown. However, if user inputs “cheese pizzas” or “cheeze pizza”(any other form of typo) the entity is still correctly identified as pizza but I am unable to find a solution to map it exactly back to the dictionary’s value. A simple but not elegant fix is to include different versions of the same phrase in the dictionary to map it to “$12 per pizza”. Thus, I would like to know if there are any methods avaliable to fix this mapping a entity to its desired value issue ?

This seems like a good use case for synonyms: NLU Training Data

1 Like

Hi @alexyuwen , Wow great. I think this looks like something that will work. Will try this out !

Thanks for the tip :smiley_cat: