ComplaintBuddy - The Complaint Management Rasa Bot

File Structure

├── actions
│   ├── actions.py
│   └── __init__.py
├── config.yml
├── credentials.yml
├── data
│   ├── nlu.yml
│   ├── rules.yml
│   ├── stories.yml
│   ├── testing_intents.yaml
│   └── training_intents.yaml
├── db_sqlite.py
├── document.odt
├── domain.yml
├── endpoints.yml
├── index.html
├── models
├── Rasa Core Visualisation.pdf
├── readme.md
├── requirements.txt
├── results
└── tests

Customer Complaint Intents Handled by the NLU model

card_not_working
card_payment_not_recognised
card_swallowed
compromised_card
declined_transfer
unable_to_verify_identity

Flow of the Rasa Chatbot

-The user initiates the Conversation with a greeting (example: ‘Hi’).
-The chatbot initiates the Activalte User Info Form rule:
-The chatbot uses utter_greet response to greet the user
-The chatbot then activates the user_info_form form to collect the following information from the user

account_no: text
first_name: text
last_name: text
email: text
phone_no: text

• The Submit User Info Form Rule is initiated to submit the information entered by the user.
• The chatbot uses the utter_submit response to respond to the user with a message(“Thank you for your response.”)
• The chatbot uses the utter_slots_values response to respond to the user with a message(“Hi, {name}, how can I help you today?”)
• The chatbot uses action_listen to get the response from the user
• The response of the user is classified with an intent, and based in the recognised intent, the corresponding rule gets triggered.

e.g.

- rule: Handle Card Not Working Complaints
  steps:
  - intent: card_not_working
  - action: action_record_text
  - action: utter_sorry_for_inconvenience
  - action: utter_card_not_working_faq
  - action: utter_has_your_problem_been_resolved
  - action: user_response_form
  - active_loop: user_response_form

This rule handles the responses for which the identified intent is card_not_working. In this case, the chatbot calls the action_record_text custom action, which sets the slot – complaint with the complaint_text entered by the user. Next, the rule uses the utter_sorry_for_inconvenience and displays utter_card_not_working_faq and then asks the user for confirmation if the problem being faced by the user was resolved by the FAQs stated by the chatbot using utter_has_your_problem_been_resolved.

The user_response_form is initiated to colllect the response from the user in either ‘yes’ or ‘no’. The response is validated using a validate_user_response_form custom action.

The action_respond_back custom action checks for the value of the response and if the response of the user was:

- ‘yes’
  - dispatch response ‘utter_submit_complaint’
	- dispatch response ‘utter_goodbye’
- ‘no’
	- insert the fields into a sqlite database:
		- account_no
		- first_name
		- last_name
		- phone_no
		- email
		- complaint
		- date
		- status
	- dispatch message: ‘utter_complaint_submission_status’
  - dispatch message: ‘utter_goodbye’  
  - use twilio client to send a message to the user indicating success in registering the complaint`

Running ComplaintBuddy

• cd complaint_buddy
• Use rasa train to train a model based on intents and examples mentioned in the nlu.yml file
• Open a new terminal and start the actions server rasa run actions
• Go to the other terminal and execute rasa shell to start interacting with the bot in the terminal window.

Starting a conversation with ComplaintBuddy

- Start a conversation by greeting the chatbot(e.g. ‘Hi’, ‘Hello’ etc.)
- Provide the data as requested by the bot
- account_no (regex: ‘[0-9]{9-18}’)
- first_name (should not be null)
- last_name (should not be null)
- email (regex: ‘^[a-zA-Z0-9.!#$%&’*+/=?^_`{|}~-]+@[a-zA-Z0-9-]+(?:\.[a-zA-Z0-9-]+)*$’)
- phone_no (regex: ‘^(\+91[\-\s]?)?[0]?(91)?[789]\d{9}$’)