Train NLU for different different meaning of a word

Rasa Open Source
1

Hi Team,

I have a requirement for unsubscription a service like user will say stop/stop service/stop messaging me… then i have to pass this to NLU for intent classification. Then NLU will understand user wants to unsubscribe a service,then it will perform the required action for subscription.

Now my question is , how can I train NLU for understanding stop a service and stop kidding has different meaning ?

I am using tensorflow_embedding ml pipeline.

unsubscribe intent:-

intent:optout

  • I want to unsubscribe
  • opt out
  • optout
  • stop
  • please unsubscribe me
  • i want to stop my service
  • stop messaging
  • unsubscribe
  • i want to stop getting messages

thanks, sandeep

2

I would suggest using the CRF here and tagging certain words that imply a service vs the user’s smalltalk.

Then on the Rasa core or your dialog manager side, branch accordingly to the entities that were returned

3

Hi Shail

As suggested I used the CRF. My requirement is similar to Sandeep but I want to extract the regex pattern name for the matched input text. Kindly refer ( Regex Name Extraction from custom Action )

I need to extract regex acc1 or acc2 according to input text

4

hey @sandeepkumarsahoo have got any solution to isuee ? please let me know how did you do it?