Document Understanding in UiPath Studio

Help Activities
, , , ,
1

Hey, I have a question related to Document Understanding. I am extracting a table from a PDF, and the number of rows may vary. I get extra rows in the results. For example, the table has 2 rows, but it returns 7 rows. The other 5 come from different sections of the document. I re-indicated the table, but the result is the same on 3 different documents. Can someone shed light on it?

Plus, which OCR is better for DU? I am using UiPath Document OCR. I was thinking about to use Omnipage OCR. Thanks a lot.

2

@Yasir_Yaqoob

You can use document understanding tab to create ML skill by training multiple documents (at least 10).
Once you train your ML skill it will capture only desired rows. Hope all pdf have same structure and rows can be any number.

Crate table fields and click on predict button it will predict it like below.
In case something indicated wrongly, you can update the fields

A software interface is displaying the extracted details of an invoice, including description, taxable value, tax amounts, and total amounts both in tabular form and as a scanned image. (Captioned by AI)
A software interface is displaying the extracted details of an invoice, including description, taxable value, tax amounts, and total amounts both in tabular form and as a scanned image. (Captioned by AI)1910×921 269 KB

3

@Yasir_Yaqoob

  1. Is it having same number of columns and same column names?
  2. train more sample to get accurate date
  3. while annotating make sure you indicate the pages from where you need data only

cheers

4

You are speaking about training first. I am not a training model. But I would like to do. Could you let me know how can I use that trained model in UiPath Studio?

5

@Yasir_Yaqoob Sure

click on create project

This image shows the UiPath Document Understanding interface, highlighting the "Create project" option for initiating a new custom model project. (Captioned by AI)
This image shows the UiPath Document Understanding interface, highlighting the "Create project" option for initiating a new custom model project. (Captioned by AI)1912×455 24.2 KB

Create a document type for example invoices

A user interface displaying a list of document types including Form1040, Invoices, Receipts, and VehicleDamageReport, with details about extraction methods and page counts, and a button to create a new document type. (Captioned by AI)
A user interface displaying a list of document types including Form1040, Invoices, Receipts, and VehicleDamageReport, with details about extraction methods and page counts, and a button to create a new document type. (Captioned by AI)1932×452 36.4 KB

click open document type

The image shows the 'Invoices' section of the DUAdvanceAssignment project in a document management interface, displaying details of the document type, dataset status, and a button to open the document type. (Captioned by AI)
The image shows the 'Invoices' section of the DUAdvanceAssignment project in a document management interface, displaying details of the document type, dataset status, and a button to open the document type. (Captioned by AI)1925×435 37.9 KB

import at least 10 files and indicate the fields

A screenshot of UiPath Document Understanding interface displaying an imported invoice with various tax breakdowns and totals, alongside a table summarizing taxable values and tax amounts. (Captioned by AI)
A screenshot of UiPath Document Understanding interface displaying an imported invoice with various tax breakdowns and totals, alongside a table summarizing taxable values and tax amounts. (Captioned by AI)1917×840 258 KB

6

I knew that part. Next, If I want to build automation in UiPath Studio, how will I select that trained project?

7

@Yasir_Yaqoob you have to create ML skill for extraction with automated training option

This image shows a user interface for managing extractors in a project, highlighting options to create a new extractor and displaying existing extractor details such as name, type, document type, status, and creation date. (Captioned by AI)
This image shows a user interface for managing extractors in a project, highlighting options to create a new extractor and displaying existing extractor details such as name, type, document type, status, and creation date. (Captioned by AI)1918×488 49 KB

The image shows a user interface for training an extraction dataset, with fields for extractor name, document type, model, and version, and an option to use GPU. (Captioned by AI)
The image shows a user interface for training an extraction dataset, with fields for extractor name, document type, model, and version, and an option to use GPU. (Captioned by AI)493×576 17.5 KB

once training is completed use this ML skill in your project with ML extractor
image

8

OK, thank you. I will give it a try very soon and will mark your answer as solution. Appreciate your time.