web
You’re offline. This is a read only version of the page.
close
Skip to main content

Announcements

News and Announcements icon
Community site session details

Community site session details

Session Id :
Power Platform Community / Forums / Power Automate / AI Builder: Classify t...
Power Automate
Unanswered

AI Builder: Classify text into custom categories | Increase performance

(0) ShareShare
ReportReport
Posted on by 99

Hi All,

 

Let me know your inputs on how to increase performance score for the AI builder custom model "Classify texts into custom categories". I fed around 200 data into the source data table. My current version of the model has a performance score of 55 which is not enough to proceed to use the model. I'll be adding more data to the data table. Please let me know any tips that helped you in increasing the performance score of this model

Categories:
I have the same question (0)
  • plarrue Profile Picture
    Microsoft Employee on at

    Hi @arun_kj ,

     

    One option will be to create your custom prompt with your own data in a prompt to classify your text into a given categories, so no need to build or train a model.

    Use your own data in a prompt | Microsoft Learn

     

    Regards,

    Philippe

  • arun_kj Profile Picture
    99 on at

    Hi @plarrue ,

     

    I have tried that as well. However I was not clear on how to use data part in the prompt. Could you please give me the right idea? The attached image shows the prompt I created. Wrt Data part, I have a Dataverse table which already has some data on categorization. Like around 200 mail bodies and the corresponding categorization already done. This is selected in the 'Data used' section. How can I use the Filter attribute and Filter Value fields to help the prompt to provide better predictions.

    arun_kj_0-1717599576932.png

    Regards,

    Arun K J

     

  • plarrue Profile Picture
    Microsoft Employee on at

    Hi @arun_kj ,

     

    Happy to read that you want to explore prompts for your scenario.

    I believe Email Classification2 is your Dataverse table with data, so in your prompt you'll need to specify that you want to classify the input text into of the input categories from the <Table>.

     

    Some Docs and a learning module that can help you

    Use your own data in a prompt | Microsoft Learn

    Create AI Builder grounded prompts with Dataverse data - Training | Microsoft Learn

     

    Regards,

    Philippe

  • arun_kj Profile Picture
    99 on at

    Okay so to clarify, the table data we give in prompt is used only for getting reference. It is not used to train the prompt like how dataverse table data is used to train an AI model. Is that right?

     

    Regards,

    Arun K J

  • plarrue Profile Picture
    Microsoft Employee on at

    The Dataverse table in your prompt is referenced as a data source.

    There is not any training model process in the prompt.

     

    here's some example of scenarios

     

    • Classify the Email into one of these Category.Name matching based on Category.Description.

     

    Use your own data in a prompt | Microsoft Learn

  • arun_kj Profile Picture
    99 on at

    Hi @plarrue , I tried creating a prompt in the way you suggested but I'm getting error while testing the prompt. Kindly have a check if you're familiar with it.

    arun_kj_0-1719468708907.pngarun_kj_1-1719468718579.png

     

Under review

Thank you for your reply! To ensure a great experience for everyone, your content is awaiting approval by our Community Managers. Please check back later.

Helpful resources

Quick Links

Introducing the 2026 Season 1 community Super Users

Congratulations to our 2026 Super Users!

Kudos to our 2025 Community Spotlight Honorees

Congratulations to our 2025 community superstars!

Congratulations to the April Top 10 Community Leaders!

These are the community rock stars!

Leaderboard > Power Automate

#1
Vish WR Profile Picture

Vish WR 785

#2
Valantis Profile Picture

Valantis 669

#3
Haque Profile Picture

Haque 563

Last 30 days Overall leaderboard