Review workflow: Fine-tuning predictions

Once Impira extracts data from your files, it's important to check our work to ensure we're on the right track. Any and all corrections and confirmations teaches Impira to make more accurate predictions.

The review workflow is an easy way to go over low confidence predictions. It's set up like a queue and using it is as simple as going down a checklist.

How to use the review workflow

Enter the review workflow using the Review predictions button in the top-right corner of the Collection view. The Review predictions button will be automatically highlighted in black if you have any red ”review recommended” markers appear within your Collection.

  1. Click Review predictions to open the review workflow window so you can quickly verify or correct predictions.
  1. Make sure the blue box is highlighting the correct data on your document and is depicted correctly in the Value box on the right.
  1. Make any appropriate adjustments to the blue box or Value box, and when everything looks right, choose Confirm value and highlighted area.
Note: Because Impira continuously learns from every value that you correct or verify, you’ll see the improvements in real time. Other predictions that have been marked with a red ”review recommended” marker at first may update with a green “high confidence” marker and be removed from the review queue as you work.

Seeing green “high confidence” markers or black “manually extracted” markers across all your cells means that Impira has correctly extracted data from your documents and your new dataset is ready for you to use.

You can do things like export your data as a CSV file, or read your data via Impira Read API.

Learn more about how we use visual markers to indicate machine learning confidence.

Sorting the review workflow

By default, the review workflow is sorted so that the predictions with the lowest confidences show up first. This is generally the most efficient way to review your values and improve the accuracy of your predictions.

However, you can also sort your predictions by field or by file order. These options can be useful if having context about other valid field values is valuable or you want to ensure that your data is correct on a row-by-row basis.

© 2023 Impira Inc. All rights reserved. This site is built with Motif.