Exercise 1: Extracting data from structured documents

Step 1: Install the Required Package

• Open UiPath Studio and create a new Blank Project named "Extraction Process".
• Click Manage Packages in the top toolbar.
• Locate UiPath.IntelligentOCR.Activities and install it.

Step 2: Create the Data Folder Structure

• Open File Explorer and navigate to your project directory: Documents > UiPath > Extraction Process.
• Inside the project folder, create a new folder named Data.
• Inside the Data folder, create two subfolders: Inputs and Outputs.
• Place all input documents (PDFs to be processed) into the Inputs folder.

Step 3: Configure the Taxonomy Using Taxonomy Manager

• In UiPath Studio, click Taxonomy Manager in the top toolbar.
• Create a new Group and name it IRS.
• Under the IRS group, add a Category named Forms.
• Under the Forms category, add a Document Type named Form1040X.
• Under Form1040X, add the following Fields:
  • FirstName — type: Text
  • LastName — type: Text
  • SocialSecurityNumber — type: Text
  • Address Line 1 — type: Text
  • Apartment Number — type: Text
  • City Town Zip — type: Text
  • Return_2019, Return_2018, Return_2017, Return_2016 — type: Boolean (one entry per year)
  • Dependents — type: Table. Inside the Dependents table field, add columns:
    • F_Name (type: Text)
    • SSN_Number (type: Text)
  • Signature — type: Boolean
  • Date — type: Date
• Create a second Document Type named VehicleRegistrationForm under a new Group called Automobiles, Category Forms. Add the following fields:
  • Registration-number — type: Text
  • Inspector-name — type: Name
  • Vehicle-license-number — type: Text
  • Driving-license-number — type: Text
  • Cost-of-Damage — type: Number

Step 4: Set Up the Main Sequence in Main.xaml

• In UiPath Studio, open Main.xaml.
• Add a Multiple Assign activity as the first activity in the Main Sequence.
• Configure the following variable assignments:
  • InputsFolder = ".\Data\Inputs"
  • OutputsFolder = ".\Data\Outputs"
  • AllFiles = Directory.GetFiles(InputsFolder)
• In the Variables panel (Data Manager tab), confirm the following variables are created:
  • AllFiles — type: String[], scope: Main
  • OutputsFolder — type: String, scope: Main
  • InputsFolder — type: String, scope: Main

Step 5: Add the Load Taxonomy Activity

• Below the Multiple Assign, add a Taxonomy container (scope activity).
• Inside the Taxonomy scope, add a Log Message activity with message: "** Load Taxonomy **", Log Level: Info.
• Below the Log Message, add a Load Taxonomy activity.
  • In the Output field, assign the result to a new variable: DocTaxonomy (type: DocumentTaxonomy, scope: Main).

Step 6: Add a For Each Loop to Iterate Over All Files

• Below the Taxonomy scope, add a For Each activity.
  • Set In to AllFiles.
  • In the Properties panel, set Argument type to String.
• Inside the For Each Body, add a Log Message with the message expression: "Processing File : " + currentText, Log Level: Info.

Step 7: Add the Digitize Document Activity

• Inside the For Each Body, after the Log Message, add a Digitize Document activity (from Activities panel under App Integration > Document Understanding).
  • Set Document Path to currentText.
  • Set OCR Engine to UiPath Document OCR.
  • In Output, assign:
    • Document Text → new variable DocText (type: String, scope: For Each currentText)
    • Document Object Model → new variable DocDOM (type: Document, scope: For Each currentText)

Step 8: Add the Classify Document Scope

• After the Digitize Document activity, add a Log Message with message "** Classification **".
• Then add a Classify Document Scope activity.
  • Set Document Path to currentText.
  • Set Document Text to DocText.
  • Set Document Object Model to DocDOM.
  • Set Taxonomy to DocTaxonomy.
  • In Output, assign Classification Results to a new variable DocClassificationResults (type: ClassificationResult[], scope: For Each currentText).
• Inside the Classify Document Scope, drop in an Intelligent Keyword Classifier from the Classifiers section of the Activities panel.

Step 9: Configure the Intelligent Keyword Classifier – Manage Learning

• Click on the Intelligent Keyword Classifier activity in the canvas.
  • In Properties, set LearningFilePath: point to a file in the DocumentProcessing folder (e.g., "DocumentProcessing\LearningFilePath") — save this file via the file browser dialog.
• Note: Create a blank new file LearningFilePath.json in the DocumentProcessing folder.
• Click Manage Learning inside the activity.
  • In the Manage Learning dialog, you will see Document Types: Form1040X and VehicleRegistrationForm, each showing "No training for this document type."
• For Form1040X, click Start Training…:
  • Select 3–4 PDF files from Data\Inputs.
  • OCR Engine is pre-populated: UiPath Document OCR, with the endpoint and API key already set.
  • Click Start Training and wait for the digitization and training progress bar to complete (100%).
• For VehicleRegistrationForm, click Start Training…:
  • Select 3–4 PDF files from Data\Inputs.
  • Click Start Training and wait for completion.
• After training completes, both document types will show their trained word vectors in the Manage Learning dialog.
• Click Save and close the Manage Learning dialog.

Step 10: Add the Present Classification Station

• After the Classify Document Scope, add a Log Message with message "** Classification Validation **".
• Add a Present Classification Station activity.
  • Set Document Path to currentText.
  • Set Document Text to DocText.
  • Set Document Object Model to DocDOM.
  • Set Taxonomy to DocTaxonomy.
  • Set Automatic Classification Results to DocClassificationResults.
  • In Output, assign Validated Classification Results to a new variable DocClassificationResults_H (type: ClassificationResult[]).

Step 11: Add the Train Classifiers Scope

• After the Present Classification Station, add a Log Message with message "** Classification Trainer **".
• Add a Train Classifiers Scope activity.
  • Set Document Path to currentText.
  • Set Document Text to DocText.
  • Set Document Object Model to DocDOM.
  • Set Taxonomy to DocTaxonomy.
  • Set Human Validated Classification Data to DocClassificationResults_H.
• Inside the Train Classifiers Scope, drop in the Intelligent Keyword Classifier Trainer activity.
  • Set Learning File Path to "DocumentProcessing\..." (same path used in Step 9).
  • Click Configure Classifiers to verify the classifier is bound correctly.

Step 12: Run and Validate the Process

• Click Debug (or Run) in the top toolbar to execute the automation.
• Monitor the Output panel to verify the process logs:
  • ** Load Taxonomy **
  • Processing File : .\Data\Inputs\<filename>
  • ** Classification **
  • ** Classification Validation **
  • ** Classification Trainer **
• Confirm that each file in the Inputs folder is processed through Digitize → Classify → Validate → Train in sequence.
• The automation processes all files in AllFiles (Form1040X PDFs and VehicleDamageReport PDFs) iteratively, training the classifier incrementally with human-validated results.

Step 1: Add a For Each Loop for Classification Results

• After the Train Classifiers Scope activity in Main.xaml, add a For Each activity.
  • Set In to DocClassificationResults_H.
  • Rename the iterator variable to currentClassificationResult.
  • In the Properties panel, set Item name to currentClassificationResult and Argument type to UiPath.DocumentProcessing.Contracts.Results.ClassificationResult.
• In the Variables panel, confirm a new variable Counter is created (type: Int32, scope: For Each currentText).

Step 2: Add the Data Extraction Scope

• Inside the For Each currentClassificationResult Body, add a Log Message with the message "** Extract Data from Document **", Log Level: Info.
• Below it, add a Data Extraction Scope activity (from Activities > App Integration > Document Understanding).
  • Configure the following inputs in the Properties panel:
    • ClassificationResult: currentClassificationResult
    • DocumentObjectModel: DocDOM
    • DocumentPath: currentText
    • DocumentText: DocText
    • Taxonomy: DocTaxonomy
  • In Output, assign Extraction Results to a new variable DocExtractionResults (type: ExtractionResult, scope: For Each currentClassificationResult).
• Inside the Data Extraction Scope, drop in a Form Extractor activity (from Activities > Extractors).

Step 3: Configure the Form Extractor Template (Template Manager)

• Click on the Form Extractor activity, then click Manage Templates (or open Template Manager from the activity).
• In the Template Manager dialog, click Create Template.
• In the Create a new template dialog:
  • Set Document Type to Form1040X.
  • Set Template name to Template_1040X.
  • Select one file from Data\Inputs folder (preferably a native document).
  • Set OCR Engine to UiPath Document OCR.
  • Pre-fill Endpoint, Api Key, and Timeout fields with the same credentials used previously.
  • Leave Apply OCR on PDF as Auto.
  • Click Configure — the system digitizes the document and opens the Template Editor.
• In the Template Editor, the document opens showing the Form 1040-X (two pages). Map the fields as described in the video.
• After all fields are mapped, click Submit to save the template.

Step 4: Configure the Regex Based Extractor

• In the Data Extraction Scope, also drop in a Regex Based Extractor activity (from Activities > Extractors).
• Click Configure Regular Expressions inside the activity.

• Click Save in the Configure Regular Expressions dialog.

Step 5: Configure Extractors via Configure Extractors Dialog

• Inside the Data Extraction Scope, click Configure Extractors.
• In the Configure Extractors dialog:
  • For Form1040X: check the Form Extractor checkbox (blue check) to activate it.
  • For VehicleRegistrationForm: leave Form Extractor unchecked; check Regex Based Extractor.
• Click Save.

Step 6: Add the Present Validation Station

• After the Data Extraction Scope, add a Log Message with the message "** Validate Extracted Data from Document **", Log Level: Info.
• Below it, add a Present Validation Station activity.
  • Configure inputs:
    • Document Path: currentText
    • Document Text: DocText
    • Document Object Model: DocDOM
    • Taxonomy: DocTaxonomy
    • Automatic Extraction Results: DocExtractionResults
  • In Output, assign Validated Extraction Results to a new variable DocExtractionsResults_H (type: ExtractionResult, scope: For Each currentClassificationResult).

Step 7: Build the Results File Path

• After the Present Validation Station, add a Sequence container named Get ResultsFilePath.
• Inside the sequence, add a Log Message with the message "** Create Path for Results File **", Log Level: Info.
• Below it, add a Multiple Assign activity with the following assignments:
  • First row:
    FileName = Path.GetFileNameWithoutExtension(currentText)
    Create a new variable FileName (type: String, scope: For Each currentClassificationResult).
  • Second row:
    ResultsFilePath = Path.Combine(OutputsFolder, FileName + "-" + Counter.ToString("000") + ".xlsx")
    Create a new variable ResultsFilePath (type: String).

Step 8: Add Export Extraction Results and Output Loop

• After the Get ResultsFilePath sequence, add a Log Message with the message "** Extract Data/ Outputs **", Log Level: Info.
• Below it, add an Export Extraction Results activity.
  • Set Extraction Results input to DocExtractionsResults_H.
  • In Output, assign DataSet to a new variable DocExtrcationResultsDS (type: DataSet).
• Below it, add a For Each activity named For Each currentDataTable:
  • Set In to DocExtrcationResultsDS.Tables.
  • Rename iterator to currentDataTable.
  • Add a Write Range Workbook activity:
    • FilePath = ResultsFilePath
    • Table Name = currentDataTable.TableName
    • Table = currentDataTable

Step 9: Save and Verify the Final Workflow

• Press Ctrl+S to save Main.xaml.
• The overall Main Sequence flow is now: Multiple Assign → Taxonomy (Load Taxonomy) → For Each currentText [Digitize → Classify → Present Classification Station → Train Classifiers → For Each currentClassificationResult [Extract → Validate → Build Path → Export]].
• Click Debug to run the full process and confirm all steps execute without errors and the output is stored in an Excel file in the Outputs folder.