Document Understanding Release Notes
2022.10.0
Stay up to date with all the latest news regarding Document Manager by going through the next list of changes that have occurred since the last LTS release until now.
Dataset Diagnostics is a new feature that helps you build effective datasets by providing visual and written guidance throughout the whole process of training a new model.
Schema import can now be applied for multi-value fields. You can also export or download a schema with multi value fields.
Please be aware that multi valued fields are compatible only with the models that have the version 2022.10 or higher.
Checkboxes can be used with multi value fields and be applied in different forms. For example, you can use of of the following checkbox options: Label the options; Label the checkboxes, with separate field for each checkbox; Label only the checkboxes with a single Multi-valued field; or Label the options with a single Multi-valued field.
The Import function now has an Import Status screen that offers details about the process, such as Import type, used OCR engine, Processed documents, Skipped duplicated, or Errors. You can choose to view the logs for more details, or to close the dialog box.
Improvements have been made on the labelling of multi value fields. You can now see in the default collapsed state of the multi value field two values, and all labelled values in the expanded state.
You can now use in Document Manager the Remove project option that is nested in an action menu and allows you to delete a previously created project.
Importing a dataset from Validation Station is now done without having the fields manual edited. If a manual edit is needed, the user can do this after importing the dataset.
The OCR settings have been enhanced, meaning that once an OCR engine is set, it can be used for a Document Manager session.
Improvements have been made on the Create new document type Semi-Structured AI window, allowing you to select an out-of-the-box pre-trained document type and to customize it as needed. This option estimated the dataset size needed for training based on the number of fields, languages and layouts.
The Create new project popup box has been redesigned and brings an Advanced options section. Here you can choose to import and select a project from AI Center by using the Import from AI Center and the AI Center projects options.
Export dialog now has an additional option "All" for exporting all data, including unlabelled data. This allows you to migrate an entire document type collection from one environment to another, which was not possible before.
The Export files popup box has a new button available, the Download to Excel option, that allows you to download the exported files in an Excel format.
The Export button from Document Manager has been renamed as Export to AI Center.
The Savebutton has been renamed as Create in the dialog box of creating a Document Manager session.
Improvements have been made in the digitizing process at the import time, the same results being received, no matter of the used importing procedure (on premises or Studio desktop).
- Fixed a bug occurring when files were imported in Document Manager, resulting in a
List index out of rangeerror. - Fixed a bug that was causing import errors in Document Manager due to naming issues.
- Fixed a bug that occurred when labelled documents were exported from Document Manager. The bug was fixed and the labelled documents are now exported without problems.
- Fixed a bug occurring when running an evaluation pipeline on a model trained with the special
line_detection mode, causing predictions to be different than when called from the ML skill.
{ symbol.
When you first access a Document Manager session in an AI Center Standalone offline environment, you are redirected to a service not yet enabled on the environment. The same thing can occur when a session expires in an Automation Suite offline environment.
