When train the data at a time get None value

Rasa Open Source
1

when train my data first time it is take slot when use second time slot is not getting intenet it is by default get with ‘None’ here is the my chat history: Chat History

Bot You

qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq 1 action_listen qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq 2 hello intent: greet 0.97 qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq 3 utter_greet 1.00 Hello! How can I help? action_listen 1.00 qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq 4 looking for admission intent: inform 0.80 qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq 5 utter_ask_location In what location? action_listen 1.00 qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq 6 india intent: inform 0.95 qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq 7 slot{“location”: “india”} utter_ask_university In which University?

Current slots: course: None, location: india, matches: None, university: None

? The bot wants to run ‘action_listen’, correct? Yes ? Next user input (Ctr-c to abort): IIT ? Is the NLU classification for ‘IIT’ with intent ‘None’ correct? (Y/n)

2

Hi @dipenpatel235, from what I have seen in other questions here on the forum (such as here), getting None intents can happen for instance when you type an unknown word (such as ‘IIT’ in your case). What pipeline are you using? As a rule of thumb, if you want your chatbot to correctly handle unseen words, I recommend using CountVectorsFeaturizer with the OOV (out-of-vocabulary word) functionality in your pipeline, more on it for example in this thread.