Hello guys! Need Help!
In below image, there is a database of rasa
My question is when the intent is goodbye with 99% confidence, why it is not replyed by a bot , it reply with another action
Hi @yugalahir, this depends entirely on how your stories are written, as well as what policies you are using. I see you’re using only the TEDPolicy - is there a reason you’re not using the RulePolicy or MemoizationPolicy as well?
Thank you for your reply
When I am using TEDPolicy with RulePolicy some stories results are not appropriate, that’s why I remove the rule policy, please suggest which policy to use
Interesting. RulePolicy makes predictions entirely based on the rules you set for it. Maybe this inappropriate behaviour stems from a bad or over-generalized rule, and not because of the RulePolicy. Are you using any rules at all?
hm yeah, i’d be interested to see your stories/rules and the issues you were having with the RulePolicy. Because the RulePolicy is the recommended policy for single turn interactions like goodbye
snek yes using 2 rules
@akelad ok will share, as I developed the WhatsApp chatbot using rasa.
It is misclassification solved by using
1 MemoizationPolicy
2 TEDPolicy (max_history: 5 , epochs: 150)
BUT RULES NOT WORKING
**RULES.yml**
rule: Say goodbye anytime the user says goodbye
steps:
intent: goodbye
action: utter_goodbye
rule: Say ‘I am a bot’ anytime the user challenges
steps:
Do you have the RulePolicy specified in the config file for this case as well? And also your rules.yml should start with a rules: entry, just like in your stories file