Document Understandng

Anyone has sample project to understand document understanding with sample files ?
Thanks in advance

Hi @Sakshi_Jain

Did u check out some video in YouTube regarding document understanding?

You can check with @Lahiru.Fernando he can help you

Hi, go to:
https://academy.uipath.com/

Then search for: UiPath Document Understanding Overview

There, during the course, a sample DU workflow is created / can be downloaded.

@Adrian_Star @NIVED_NAMBIAR Yes did watch some videos and tried on system

How do we decide which type of activities we need use for these
load taxonomy,digitize document ,classify,extract,validate,export

Different types of OCR :tesrat,omnipade,microsoft,uipath,google
text extractor : regex based extrator, form extrator, intelligent form extrator, Ml extrators for invoice, receipts, purchase orders etc
Classifier : keyowrd based …

also it was not working for set of images i was trying to get the data

can see the code of the activities we are doing , like if i want same automation through python anyway i can get the code ?

The taxonomy is used to recognize the documents and data you want to extract and must always be prepared for each document that is subject to the process.

Digitization is the conversion of a scan to a digital version of a document. An attempt is made to read characters and assign their positions in the digital version of the document, so that, for example, at the validation station, it is possible to indicate the position, and the data extraction process understands it.

The different types of data extraction depend on the degree of precision you want to achieve.
I mean, are you satisfied with the 60% reading efficiency or maybe 90% +?

If the value you want is consistently between some text and the document being read to an editable PDF then you can use REGEX.
If you are dealing with a scan, you are using OCR engines.
OCR engines have different efficiency and possibilities (scaling, rotating). It is important that they support the language for which you want to read the data.

There are OCR engines that automatically rotate a skewed image, others don’t.

If you have a Document Understanding server, then you can train the ML model to improve its effectiveness. Without a learning model, performance may vary depending on the quality of the document.

As for the UiPath source code for learning, it is probably not generally available to the Developer from the flow level.

1 Like