UiPath Community 2022.4 Stable Release - Document Understanding

Greetings, UiPath Community! :dancer:

We’re delighted to announce what we’ve been up to lately and present you our latest release with various delighting features and improvements! :dancer:t2: Keep on reading to find out more!

Consume Forms AI Extraction Models using the ML Extractor :mag_right:

Use AI to extract data easier than ever! With this release, you are able to consume the custom configured Forms AI Extraction Models using the ML Extractor - simply use the endpoint of the Forms AI model with the ML Extractor (just like you would do for a public endpoint) and you will be able to extract the data just as trained in the Forms AI session.

Detect handwriting using UiPath Document OCR :writing_hand:

Our own UiPath Document OCR is now able to detect and ready handwritten text! As a consequence, we will deprecate the Intelligent Form Extractor and suggest you to use the Form Extractor instead, which is now enhanced with Signature Detection capabilities - so that you won’t miss the IFE :slight_smile:

Improved Digitization Algorithm :wrench:

We have been working hard to improve the digitization algorithm for increased performance and more accurate results! In this sense, we have enabled the PDF and ML Packages to use the same digitization library as the Digitize Activity, as well as enhanced this with support for parsing native PDFs. Shall you encounter different behaviour between the packages, please do your feedback :slight_smile:

Apply OCR on a need-to basis :control_knobs:

Ever wanted to process native PDFs, without being required to go through OCR? You can now do this by using the Apply OCR on PDF property of the Digitize Activity, which now allows you to use one of 3 configuration options:

  • auto (default): let the algorithm decide whether OCR is required or not
  • yes: always apply OCR
  • no: when PDFs are passed as inputs, do not apply OCR

Accessibility Improvements of the Validation & Classification Station :medal_military:

Enjoy now accessibility improvements in our validation tools by making interactive elements keyboard accessible - so that you are now able to easier & quicker validate your documents!

Train Public Datasets using the ML Extractor & Classifier Trainers :man_lifting_weights:

With this release, both the ML Extractor Trainer and the ML Classifier Trainer can be used to train public datasets: this means that, one can provide the Activity a URL referencing a public Dataset as configured in AI Center, where the data can be sent. This is particularly helpful, when the robot is not deployed on the same environment as AI Center and the Orchestrator. Find more details in the specific documentation.

Continuing our efforts for the .net 5 upgrade :arrow_double_up:

We have adapted our packages to support .net5 core framework:
PDF :white_check_mark:
Omnipage :white_check_mark:
Abbyy :white_check_mark:

We are also providing .net5 cross platform support for:
IntelligentOCR :white_check_mark:

We hope this feature are useful to you - and if you have any feedback, please give us a shout - we’re always happy to receive it! :star_struck:

8 Likes

Excited by these updates! Thanks to the DU team at UiPath.

1 Like

I have shared a post on UIPath Forum regarding the problems that I was facing while working on document understanding but got no solution.

So here I am attaching the link of the post and if anyone get the solution for this please guide me over this issue.

It would be a great Help!!

Hi Kunal, thanks for reporting! I asked my colleague @alexcabuz, PM of the handwriting capabilities, to have a look into it :slight_smile:

1 Like