UiPath Document Understanding

Document Understanding is the UiPath product that enables bots to intelligently extract, interpret, and process data from documents. These documents can be a wide range of types from handwritten, scanned images to well-structured PDFs.

Document Understanding utilizes Optical Character Recognition (OCR) and Machine Learning (ML) / Artificial Intelligence (AI) to extract text from documents, analyze the text, and then provide the desired output.

UiPath’s Document Understanding provides a selection of different ML packages that may be utilized if you’ve been pretrained by UiPath, or you have the option to custom train the packages by utilizing AI Center.

Smartbridge is an UiPath Partner


AI Center

AI Center is a product that enables you to deploy, consume, manage, and improve machine learning models. This product allows you to choose which ML package you would like to use whether it be a custom-built package, pretrained, or trainable.

In one of our recent projects here at Smartbridge, we utilized Document Understanding using the pretrained, publicly accessible endpoint for the “Invoices” ML package provided by UiPath. This Invoices package extracts relevant data from invoices such as invoice number, invoice date, due date, and total amount. Without any custom training to the package, we were able to completely automate about 30% of our invoice data to be extracted and input into an accounting system. The remaining 70% of invoices required some type of human interaction, such as selecting a missing field that was not extracted by using Validation Station. We then made the decision to utilize AI Center to improve our automation rate by custom training the package to our needs.

For more information about AI Center, check out the UiPath documentation here.

Image of currently available UiPath Document Understanding Machine Learning Packages courtesy of UiPath

Validation Station

Validation Station is an activity in UiPath used when a document requires human validation. Within the UiPath Studio code, the process can be set up to send a document to Validation Station for any condition you want to define. Usually, this condition is when a required field is missing from the extraction, or the extraction is below a certain confidence level. This activity can be implemented with Action Center within UiPath Orchestrator to pause the processing of one invoice and continue with the remainder. Then once the invoice waiting in Action Center is validated, it will complete the rest of the job.

After custom training the provided UiPath Invoices machine learning package with our own training pipelines, we were able to fully automate about 70% of all invoices without any human interaction needed. For this number to improve over time, a developer must run more pipelines with new human validated invoice data as it comes through.

For more information on Validation Station, please refer to the UiPath documentation here.

UiPath Version 2021.6

With the current state of Document Understanding and Action Center, it is not possible to fully automate the retraining of a machine learning package. A developer is required to take the human validated data, upload that data to Data Manager, and then export the data in a new format. The newly formatted data is then used by the developer to run a pipeline to train the machine learning package.

In the upcoming update, UiPath has made strides towards completing the loop of automatic, continuous retraining of the AI used to extract the data from documents. They have updated the training activity in Studio to automatically format the human validated data to be uploaded to AI Center. This data may then be used for a pipeline run which is scheduled to run on a periodic basis. This allows for the process to continually learn and improve over time with low maintenance needed.

For more info on the UiPath 2021.6 update, please refer to this forum post here.

Image of the Document Understanding framework loop courtesy of UiPath

Personally, I am very excited for the upcoming 2021.6 update as it completes the automation loop for the retraining of ML packages. The 2021.6 update, Document Understanding, and AI Center used in conjunction provide a truly robust autonomous process that learns and improves over time.

Looking for more on Automation?

Explore more automation insights and expertise here.

Keep Reading: Case Study — Real Estate Firm’s AP Invoice Automation

Originally published at https://smartbridge.com on July 29, 2021.



We’re geeks for the enterprise systems and tech that sustains and strengthens business. Simplifying business transformation. Smartbridge.com

Love podcasts or audiobooks? Learn on the go with our new app.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store