Reviewing the extracted information#

There can be different ways to access Documents in the Document Validation UI to review the information that was retrieved by the Extraction AI or to add information that wasn’t, depending on whether the application is hosted by our Konfuzio server, or it is integrated to a third party application, using their server of choice.

Document Validation UI hosted by the Konfuzio server#

First, a Document will be assigned to you to review, and you will then receive an email informing you that you have a new Document to review, which will include:

  • A link to the assigned Document on the Konfuzio app.

  • A link to the current Document Validation UI Guide “Review Documents”, for guidance during the review process, if needed.

Document Validation UI integrated with a third party application#

The Document which should be reviewed is accessed up to the third party application, who decides if it’s done by a link in an email or if it’s accessed directly from the application itself. The way in which the Document Validation UI is integrated is completely up to the end-user.

Available actions to review the Document data#


If the data was correctly extracted by the AI, you will be able to “accept” it, as shown in the clip below.

Accept Annotation

This option will be available only when hovering over the data that has not been revised yet.

Accept all extracted data in a Label Set#

You can also accept all the extracted data that has not yet been revised within a specific Label Set, instead of accepting one by one.

To know exactly what are the rows that will be accepted, you just need to hover over the “Accept all” button, which will only be available if there is unrevised data in that Label Set.

Accept all Annotations in Label Set

Mark as Missing#

You might encounter the following scenarios when you would like to mark data as missing:

  • If there was no data extracted for a defined Label.

  • If there was data extracted, but it is not correct, and there is no text in the Document that belongs to that specific Label.

Mark Annotation as Missing

As shown in the previous video, this option will only be available when hovering over an unfilled Label. Once an Annotation is marked as Missing, it won’t be editable, unless you restore it to the previous state by clicking the corresponding button.

Restore Missing Annotation

Mark all unfilled rows in a Label Set as Missing#

If there are many Labels in a Label Set which do not have any data extracted, it is possible to mark all of them as missing at once, as shown in the clip below:

Mark all empty as Missing

As with the “Accept all” button, mentioned previously, this button will also only be enabled if there are empty Labels remaining.

Edit and correct AI results#

If a specific data exists in the Document but was not extracted correctly, you will be able to edit it to make sure its value is the expected one. The edit mode gets enabled when you click on this data or the area where this data should appear.

You can select new text from the Document Page, either by creating a bounding box around the data you want to add or by clicking on the specific data, edit the data manually, and even delete (using Backspace key) the content if you want to completely replace it.

Edit Annotation

Edit by clicking on specific data

It is also possible to create a new Annotation when it was not found.

Create Annotation when editing

When a Label appears multiple times inside the same Annotation Set (group of extracted data within a specific Document), and the extracted text is entirely deleted for that Label, that specific row will be removed from the list, while the others remain.

Delete multiple Annotations

Add new data manually#

If some data was not extracted by the AI, you can manually create it by selecting the text from the Document Page, and choosing the corresponding Annotation Set (group of extracted data within a specific Document) and Label that it will belong to.

Add Annotation

Please note that, as mentioned previously, the more an Extraction AI is trained the less data will have to be manually added, edited or marked as missing.

Create Annotation Sets#

As mentioned in previous sections, an Annotation Set is a group of extracted data, and more specifically, it is a Label Set (with its corresponding Labels) extracted for a specific Document (check out our Configuration section for more information on these concepts).

It is only possible to create new Annotation Sets when a Label Set is set up to allow multiple of these groups of data, and you can find detailed information on how to do this in the following guide.

Create Annotation Set