ixp
latest
false
  • Overview
    • Introduction
    • Extracting data from unstructured documents
    • Building and deploying models
  • Model building
    • Overview
    • Uploading sample documents​
    • Reviewing predictions
    • Validating extraction predictions
    • Configuring the model
  • Model validation
  • Model deployment
  • Frequently asked questions
Important :
Communications Mining is now part of UiPath IXP. Check the Introduction in the Overview Guide for more details.
UiPath logo, featuring letters U and I in white
PREVIEW

Unstructured and complex documents user guide

Last updated Jul 2, 2025

Validating extraction predictions

The model does not currently learn directly from user annotations. To improve the performance of the model, modify your prompt instructions or model configuration.

Once you are happy with your prompt instructions, you can start validating the predictions of the model to establish ground truth for its performance evaluation. Before you do so, it is important to understand the differences between predictions and references.

Predictions and references

Predictions are the predicted values.
Note: Incorrect or missing predictions impact the model score and performance metrics in the Measure tab.
References are the matching references to the predicted values found in the document. The purpose of references is to help users find the right part of the document.
Note: Incorrect or missing references do not impact the model score or performance metrics. However, it might be helpful to add or correct references to improve the human validation experience.
There are known issues with references and highlighted spans that are being worked on. You might encounter one of the following scenarios:
  • Correct prediction and correct reference, but an incorrect span is highlighted. Usually, the first match that was found.
  • Correct prediction and correct reference, but no span is highlighted.
  • Correct prediction, but there is a missing reference or highlighted span, or both.
Note: Predictions and references are independent. This means that missing or incorrect references do not impact the predicted values.

Figure 1. Example of prediction and reference

Validating extractions

To validate extractions, follow these steps:
  1. Go to the Build tab, and select Validate predictions or any uploaded document to begin the validation process. The model highlights the relevant span of text in the document and displays the generated predictions in the side panel.
  2. Review the predictions and either confirm the extractions one by one or in bulk by confirming the entire field group. It is not recommended to select the Confirm all and next button without carefully reviewing each extraction.
  3. You can configure additional fields in your extraction schema at any point from Manage taxonomy or the Build tab.

You are recommended to validate at least 10 examples for each field, so that the model can provide meaningful validation statistics.

The validation process entails the following:
  • Confirming the extractions if they are correct.
  • Correcting the extractions if they are incorrect.
  • Marking fields as missing if they are not present in the document.
  • Configuring any additional fields in your extraction schema.
This section shows what extractions and references look like in their confirmed state (validated) and unconfirmed state (not validated).
Note: Currently, you cannot partially save your annotations in a single document.

Unconfirmed

References in the document are highlighted with a dashed border:

Figure 2. Unconfirmed reference

No extractions are marked as confirmed:

Figure 3. Unconfirmed extractions

Confirmed

References are highlighted with a solid border:

Figure 4. Confirmed reference

Extractions are marked as confirmed:

Figure 5. Confirmed extractions

Validated predictions

This section contains example scenarios of when the predictions of the model match or do not match the annotations of the user.

When the annotation matches the prediction, the extraction looks as follows:

Figure 6. Correct prediction

When the annotation does not match the prediction, the extraction looks as follows:

Figure 7. Incorrect prediction

When you added an annotation but the prediction was missing, the extractions look as follows:

Figure 8. Missing prediction

Editing extractions

In the Validate predictions page, under the Build tab, edit the extractions as follows:
  1. Select the plus icon to add a missing extraction.
  2. Select the table icon to open the table view.
  3. Confirm the extractions either from the side panel or from the table view.
  4. Hover over the extraction, and use the pencil and plus icons to correct the extractions individually:
    • The pencil icon allows you to edit the predicted value and type in the correct value.
    • The plus icon allows you to update the predicted value or reference and highlight the correct value in the document.
  5. Select Add field group to add a missing field group.

Figure 9. Validating extractions from the side panel

Figure 10. Validating extractions from the table view

Was this page helpful?

Get The Help You Need
Learning RPA - Automation Courses
UiPath Community Forum
Uipath Logo White
Trust and Security
© 2005-2025 UiPath. All rights reserved.