Lesson Guide: Extracting Tabular Data from Unstructured Documents

In this lesson, you will learn how to parse complex relational layouts—such as multi-row line-item grids nested inside invoices, receipts, or utility bills—using Noca Flow’s native Extract Table utility.

While single-value extraction tools are great for pulling out isolated data fields (like an invoice date or total amount), manual string mapping falls short when documents contain a varying number of data rows. The Extract Table node automatically scans your unstructured files, isolates table bounding regions, and transforms repetitive line-item grids into a clean data array ready for your downstream databases.

Step 1: Pinpoint Your Visual Target

Before pulling data from your file source, look at the underlying structural layout of the source asset.

1. Open your document attachment (e.g., a sample invoice PDF or utility bill statement).

2. Note the multi-row relational structure inside the bill, such as a Summary of Charges or Itemized Cost section.

3. Observe how each row contains a series of recurring columns. This is the exact data grid we will extract.

Step 2: Add the Extract Table Utility Node

Now, step onto your Noca canvas to insert our intelligent parsing layer.

1. Open your Noca Flow canvas workspace.

2. Click the central + icon located immediately after your active file source node.

3. In the application catalog action drawer, type and select Extract Table.

Step 3: Map Your Source File Variable

Direct the extraction engine to your file payload so it can begin visual parsing.

1. Click on the newly placed Extract Table Node to open its configurations panel on the right.

2. Go to your left-hand Data Kit explorer slide-out menu.

3. Select your active physical file payload variable, drag it across the canvas space, and drop it directly into the designated Target Asset input slot.

Step 4: Define the Table Column Structure

Instead of drawing complex coordinate boundaries or writing text parsing rules, simply tell Noca exactly what columns to watch for.

1. Locate the Table Column Structure parameter block inside the node properties dashboard.

2. Explicitly type out the targeted column headers you want the AI engine to isolate from your visual document matrix. For our invoice example, input:

   • Charge Item

   • Current Amount

   • Balance

3. Click Insert Node to save your configurations.

Step 5: Process and Use the Loopable Object Array

1. By setting your custom header tokens, Noca’s structural extraction layer automatically scans the layout coordinates, groups every row under your chosen headers, and formats the table into a clean data matrix.

2. If you attach a downstream node (like an automation loop step or an external spreadsheet app), you can access the extracted table array from your Data Kit. Each line item will be perfectly parsed into a clean string, allowing you to sync multi-row documents automatically!

Key Takeaways

• Automated Matrix Grouping: The Extract Table node automatically handles varying table row counts, converting visual structures into machine-readable datasets without any manual mapping.

• Header Token Optimization: Defining your target column names tells the visual AI layer exactly how to partition data fields, keeping your data structures clean and predictable.

Sensational progress! You have mastered advanced document matrix processing. You are now fully equipped to build hands-free, enterprise-grade invoicing pipelines across the Noca Flow matrix!