TechLab Part 2

1 - Duplicate existing tables

  • Duplicate the table review3031122023 and name it review3031122023_ai_analysis
  • Then, build this new table: review3031122023_ai_analysis

Image 1

2 - Analyze Airbnb comments using AI Endpoint

We will look at AI Endpoint to analyze Airbnb comments. The goal is to obtain the emotion and the main advantage from the various Airbnb comments specific to each listing. To achieve this, we need to define a language model for this purpose.

2.1 - Access the AI Endpoint website

Here you can access the AI Endpoint website: https://endpoints.ai.cloud.ovh.net/ and open gpt-oss-20b to verify the model configuration.

2.2 - Define a custom action from the DPE platform

From the data processing engine, create a custom action:

Image 2

Then upload the Python file: airbnb_paris_reviews_analysis.py

File link: https://rbnbfortechlab.s3.gra.io.cloud.ovh.net/airbnb_paris_reviews_analysis.py

Image 4

Configure the custom action:

  • Name: AI Comments Analysis
  • Function name: airbnb_paris_reviews_analysis
  • Language: Python 3.11
  • Dependencies: pandas, numpy

    • Image 3

    Now go to the Preferences tab and add the AI Endpoint API key:

    • Name: OVH_AI_ENDPOINTS_API_KEY
    • Paste the value of the key you created in your client space earlier

    Image 5

    Create the custom action and click Run. You can monitor the progress in the logs.

    Image 6

    Once finished, check the new table in Lakehouse Manager Explorer:

    Image 7

    3 - Create a dashboard to visualize the data

    To create a new dashboard, go inside the Analytics Manager.

    3.1 - Creating queries

    Create your first query, select “SQL”, and name it nb_comments_airbnb_from_december_30_to_31_2025.

    Run the first query:

    SELECT  
  COUNT(*) AS nb_airbnbs,  
  nb_comments_december_30_31_2025  
FROM (  
  SELECT  
    id,  
    COUNT(comment) AS nb_comments_december_30_31_2025  
  FROM review3031122023_ai_analysis  
  WHERE comment IS NOT NULL  
  GROUP BY id  
) sub  
GROUP BY nb_comments_december_30_31_2025  
ORDER BY nb_comments_december_30_31_2025;

    You should get something like this:

    Image 8

    Then you can create two more queries as follows:

    • nmb_comments_per_advantage_dec_30_31_2025

    Image 9

    • nmb_comments_per_emotion_dec_30_31_2025

    Image 10

    3.2 - Creating the dashboard

    You can now create a new dashboard:

    Image 11

    Add the various queries to build a complete dashboard:

    Image 12

    Demo of Superset integration can be found here: Superset tutorial

    Other useful links: