It made the performance of the NER worse, if there were mispellings it didn’t identify them, if the look up table did not contain values the NER did not recognise them as cities.
I have another question if thats ok, I keep having an error with creating FormActions now, do you have any clear examples (not the documentation) in which it is used?
I keep getting this error :
"event to log must be an instance "ValueError: event to log must be an instance of a subclass of Event.
I’m not sure how to write the story in the case, I did it in the following way:
I’m afraid that kind of makes sense, if you have a huge list of values in your lookup table then the ner_crf might overfit on that. However, if you’re willing to, you could send us your training data to hi@rasa.com and we could take a closer look
As for the FormAction, I guess you’re returning None somewhere? Could you paste the code of your form action?
I don’t think it’s the stories in this instance, I have keep changing it however nothing seems to make a difference. Do I need to call another external action or is it sufficient just to have the FormAction?
* cerca_ospedale_citta
* action_cerca_ospedale_citta
* slot{“requested_slot”: “citta”}
* inform{“citta”:“torino”}
* action_cerca_ospedale_citta
* slot{“citta”: “castellamonte”}
* ringraziamento
* utter_ringraziamento
* utter_goodbye
I don’t think it’s the stories in this instance, I have keep changing it however nothing seems to make a difference. Do I need to call another external action or is it sufficient just to have the FormAction?