Magic ETL

Magic ETL

ETL Process to Analyze Google Reviews and Generate Automated Replies

Hi everyone,

I’m exploring adding a step to an ETL process to analyze Google review data and generate recommended responses using ChatGPT or a similar tool. Currently, I have a Zapier connection set up, but it’s unreliable and inconsistent.

Here’s what I’m looking to achieve:

  1. Dataset Input: I have a dataset of Google reviews with two columns:
    • Star Rating (1–5)
    • Review Text
  2. Desired Output: I want to add a column for "Recommended Reply," where it generates an appropriate response to each review based on its star rating and text content.

Example Input Dataset:

Star Rating

Review Text

5

"Amazing service, highly recommend!"

2

"Room was not clean, very disappointed."

4

"Good experience but could improve food quality."

1

"Terrible, will not be coming back."

Desired Output:

Star Rating

Review Text

Recommended Reply

5

"Amazing service, highly recommend!"

"Thank you for your kind words! We’re thrilled you had a great experience."

2

"Room was not clean, very disappointed."

"We’re sorry to hear about your experience. Please reach out to us directly so we can make it right."

4

"Good experience but could improve food quality."

"Thank you for your feedback! We’ll work on improving our food quality."

1

"Terrible, will not be coming back."

"We apologize for your experience. Your feedback is important, and we’d like to discuss this further to resolve the issue."

Questions:

  • Has anyone implemented something similar in an ETL?
  • Is it possible to integrate ChatGPT or a comparable model (e.g., through an API or a platform like Domo) to generate the "Recommended Reply" column?

I’d appreciate any thoughts, tools, or approaches you’ve used to solve a similar problem. Thanks in advance!

Best Answer

  • Answer ✓

    Yes this can be done using DomoGPT in a Jupyter Workspace. If you come to Domopalooza, you can hear @nathankilcrease explain how this is done. He may even be willing to help out sooner than that if you ask him very nicely and offer him some legos.

    If I solved your problem, please select "yes" above

Answers

  • Answer ✓

    Yes this can be done using DomoGPT in a Jupyter Workspace. If you come to Domopalooza, you can hear @nathankilcrease explain how this is done. He may even be willing to help out sooner than that if you ask him very nicely and offer him some legos.

    If I solved your problem, please select "yes" above

Welcome!

It looks like you're new here. Members get access to exclusive content, events, rewards, and more. Sign in or register to get started.
Sign In