Some times tracker.get_slots() return as None value

Rasa Open Source
1

Hi

I’m trying custom actions in my Bot application using Python script

def run(self, dispatcher, tracker, domain): print(tracker.get_slot(‘PERSON’))

Following is my trained data

Looking for Muthu

When I type “Looking for Muthu” in Bot UI the tracker returns “Muthu” as entity/slot

But I type “Looking for Venu” the tracker returns none

It cann’t able to identify the other than Muthu as Person entity.

I’m using following configuration

pipeline:

  • name: “nlp_spacy”
  • name: “tokenizer_spacy”
  • name: “intent_entity_featurizer_regex”
  • name: “intent_featurizer_spacy”
  • name: “ner_crf”
  • name: “ner_spacy”
  • name: “ner_synonyms”
  • name: “intent_classifier_sklearn”
2

Are you training this entity extraction using CRF? This article might help

3

Thanks for the reply souvikg , As per the link I upgraded my RASA - NLU version to 0.13

And added the ## lookup: PERSON data/Person/Person.txt

Then train my data by using the following command python trainer.py train-all

But the lookup element doesn’t add in training_data.json

Do I miss anything?

4

Not sure what is the issue, you just need to add the lookup object in your training data and train a CRF with some of the examples from your lookup to create a patter of identification

"lookup_tables": [ 
                {
                    "name": "company",
                    "elements": "data/company/startups.csv"
                }
        ],
5

Hi Lakshmi,

If i am trying to print tracker.get_slot(’’) i am getting the below error TabError: inconsistent use of tabs and spaces in indentation

My exact statement looks like - print(tracker.get_slot(‘cuisine’))

Just wanted to see the values present in the slots in my customized action.

Do you have any inputs ?

Thanks.

6

It got resolved…Some issue with the tab fields in that line…Deleted and wrote again.Thanks!

7

Hi

Yes I did , Following are the steps done

1, Upgraded rasa_nlu : 0.13.5

2, Added Lookup_tables in MD file instead of json

lookup:Person

data/Person/Person.txt

3, Then updated config file as follows

pipeline:

  • name: “nlp_spacy”

  • name: “tokenizer_spacy”

  • name: “intent_entity_featurizer_regex”

  • name: “intent_featurizer_spacy”

  • name: “ner_spacy”

  • name: “ner_synonyms”

  • name: “intent_classifier_sklearn”

  • name: “ner_crf”

    features: [ [“low”, “title”, “upper”], [“bias”, “low”, “prefix5”, “prefix2”, “suffix5”, “suffix3”, “suffix2”, “upper”, “title”, “digit”, “pattern”], [“low”, “title”, “upper”] ]

Then i executed train command . My training_data.json has the following update

“lookup_tables”: [ { “name”: “Person”, “elements”: “data/Person/Person.txt” } ],

Now I’m running rasa_nlu server Still I’m getting None response for Tracker

8

is your slot_name the same as the entity name?

PERSON?

can you simply call the NLU to see if the Person entity is detected or not

9

Yes it is PERSON

print(tracker.get_slot(‘PERSON’))

It detects the PERSON entity, thus why i mentioned sometimes getting None value.

In my training data, added some examples like

Looking for Muthu

Would like to about Lakshmi

So “print(tracker.get_slot(‘PERSON’))” can able to returns Muthu or Lakshmi. If i try some other names its prints nothing

10

I will have to test the pipeline to see why this isn’t working. Will let you know

11

Hi

Can anybody have solution for this issue