Azure document Intelligence feedback mechanism with Human in the loop

Question

I'm using Robotic Process Automation (RPA) combined with Azure Document Intelligence (ADI) to automate document processing. I've built and trained a custom model in ADI using document intelligence studio. My RPA tool submits documents for processing using the ADI POST API, retrieves the extracted data using the GET API, and displays it on a user interface for verification and accuracy. To continuously improve the model's accuracy, I want to create a feedback loop where human reviewers corrects the errors and then pass this data to the model for retraining. Is there a built-in API for retraining the model based on user feedback, or how can I achieve this feedback mechanism?

Answer 1

Hi @Sai Ramesh,

Thank you for reaching out to Microsoft Q&A forum!

Azure Document Intelligence (ADI) does not currently have a built-in API for retraining the model based on user feedback.

However, there are some workarounds that you can use to achieve a feedback mechanism for your custom models. One approach is to manually update the model using the feedback data that you collect from human reviewers. You can use the feedback data to identify areas where the model is making errors, and then update the model to improve its accuracy.

To update the model, you can use the Azure Document Intelligence Studio to modify the model's training data. You can add new training data to the model to improve its accuracy based on the feedback data that you collect from human reviewers.

For more info see: Custom models accuracy and confidence scores.

Hope this helps. And, if you have any further query do let us know.

---

If this answers your query, do click Accept Answer and Yes for was this answer helpful.