Entity extraction in rasax

[Deprecated] Rasa X Community Edition
1

how to perform entity extraction with rasa x?

2

You can label entities in the training data tab :slight_smile:

3

okay fine and please update regex documentation how it is used? and if suppose i want to add regex for detecting email ids how do i specify such that it will recognize email ids through use of regex

4

Regex you still need to add to your nlu.md file for now. We’re working on adding a UI for this in future though.

If you get a chance, we’d love to get your feedback on Rasa X: Product Feedback - Rasa X

5

yes i will definately add review and if possible kindly update documentation in detail for people like me implementing for first time it would be of great help

1 Like
6

Could you let me know which part specifically you had trouble with?

7

I want to implement regex feature for verifying email on chat i have defined intents for it and now how to use regex that will verify email id for me instead of adding lines manually and another issue is i want to train my bot by integrating it to slack is it possible by importing slack conversations with rasa ui…i am using ngrok server for localhost testing and when i test my URL in event tab > request URL it does not verify i am using https:\xxxxxx.io…/webhooks/slack/webhook

8

about verifing email – i’d suggest you use duckling to extract that, it’s very robust

1 Like
9

Hi @mandarsuryavanshi, I second what Akela said about using duckling, it’s definitely the preferred way to detect emails. As for the slack connector, have you been able to connect it?

when i test my URL in event tab > request URL it does not verify i am using https:\xxxxxx.io…/webhooks/slack/webhook

If you’re still stuck on this, make sure that your rasa server is running on the port hooked up to ngrok before you attempt to verify the URL.

10

Hi @akelad, could you please help me with setting up the regex for email. Im wondering where all i have to make a change in the code. I need the bot to ask the email again if the regex doesnt match. So far

  • ive added “DucklingHTTPExtractor” in the pipeline, and ran the server
  • added a story for email collection :
## check_account
> check_account
    - utter_ask_email
* input_email{"email" : "valid@email.com"}
    - action_get_user

## check_account
> check_account
    - utter_ask_email
* input_email{"email" : "invalid@email"}
    - action_ask_email_again

How do i trigger action_ask_email_again ?

11

@agrawal-mohit the second story should be without an entity

12

I did that and entered “mohit@gmail” . It goes to action_default_fallback and action_listen instead of triggering utter_invalid_email

## check_account.invalid_email
> check_account
    - utter_ask_email
* input_email
    - utter_invalid_email

Suggestions ?

13

what nlu pipeline are you using? It’s possible that nlu isn’t recognising the intent properly

14

Got it, the problem was with the nlu. However an incorrect email mohit@gmail still sets the email slot .

15

well what does your config look like? and are you using regexes? duckling wouldn’t pick that up as an email usually