Embedding policy Feature 'entity_XX' could not be found in feature map

Rasa Open Source
1

rasa_core.featurizers - Feature ‘entity_time’ could not be found in feature map.

time is a slot and requested_sot in a form

policies:

  • name: EmbeddingPolicy

    epochs: 20

    featurizer:

    • name: FullDialogueTrackerFeaturizer

      state_featurizer:

      • name: LabelTokenizerSingleStateFeaturizer

  • name: FormPolicy

2

@azizullah2017 can you help me? anyone?

3

Can you post the content of the entities section of your domain.yml file?

4

my domain.yml don’t have entities session

5

@netcarver I have check models/dialogue/domain.yml，entities = [] can you know what wrong with me?

6

I think you need to read up on the domain.yml format and include your entities in your file.

7

@Bob Can you share your domain.yml file ?? you have not added entity_time in your domian file in a section of entity list.

8

@azizullah2017 @netcarver my train domain.yml

slots:

time:
    type: unfeaturized

when I use keraspolicy, I don’t have this problems

9 
entities:
  - entity_time

add this to your domain.yml file

10

why I need do this，can you tell me?

11

I don’t use entity_time slot

12

because

rasa_core.featurizers - Feature ‘entity_time’ could not be found in feature map.
13

But The Rasa Core dialogue engine don‘t tell me the details to do this, I also need to do that?

14

when you use embedding policy, you add entity_XX to you domain.yml even though you don’t need?

15

Could it be that you have (even accidentally) declared a core training story with the time entity?

There’s an example here. Note how the entities are declared with the intent in the example story.

If you don’t have any entities in your NLU training data, and you don’t have any in your core training stories, I don’t know what else to suggest. Your error message from core does suggest that this entity is being used somewhere.

16

I am using Embedding policy to train rasa core with stories in the format as below,

* general-greet
 - greet
* rest-inform{ENTITY_NAME: ENTITY_VALUE, ENTITY_NAME: ENTITY_VALUE}
 - rest-recommend{ENTITY_NAME: ENTITY_VALUE, ENTITY_NAME: ENTITY_VALUE}
* booking-inform{ENTITY_NAME: ENTITY_VALUE, ENTITY_NAME: ENTITY_VALUE}
 - booking-book{ENTITY_NAME: ENTITY_VALUE, ENTITY_NAME: ENTITY_VALUE}
* general-thank + general-bye
 - bye+thank

sample ENTITY_NAME - {food, name,area, post, phone number}, ENTITY_VALUE -{european, Vapiano, centre, cb13tz, 02345 65819}

The domain file format has shown below,

slots:
   entity_XX:
       type: text
entities:
- entity_XX

When I use the EmbeddingPolicy, I get the problem that the “Feature ‘entity_XX’ could not be found in feature map.” I have also tried to alter the domain file as like as below and tried to train but still ended with the above problem

slots:
   XX:
      type: text
entities:
- entity_XX

I have trained successfully with the above mentioned story format and an appropriate domain file using keras policy.

Could I get help to solve this issue?