Custom Table Column Extraction Using Prompts

Overview

Docsumo allows users to set up Custom Table Column Extraction using Custom Processing and AI-Powered Prompts. This feature enables users to define how tables should be extracted and structured using LLM prompts, reducing manual effort and ensuring accurate data extraction.

Feature Highlights

• AI-Powered Table Extraction: Use natural language prompts to define table structure and extract specific data fields.
• Customizable Column Mapping: Specify the required column names for AI-based extraction.
• Post-Processing Logic: Apply additional transformations using custom processing rules.

How to Use Custom Table Extraction

Step 1: Access Custom Processing

1. Navigate to the Custom Processing section for a table line item column in your edit field setup.

   

Step 2: Define the AI Extraction Logic

1. Use the AI_EXTRACT_TABLE function to define the table structure.
2. Provide a prompt describing the extraction logic and the required columns.

Syntax:

AI_EXTRACT_TABLE(prompt, *cols)

Example:

To categorize transactions based on description, debit, and credit fields, use:

AI_EXTRACT_TABLE("categorize transaction based on description debit and credit", "description:debit:credit")

Step 4: Save & Apply

• Click Save & Extract to finalize the setup.
• The AI model will now apply these rules automatically when processing documents.

FAQs

1. What if the AI does not extract the table correctly?

• Users can modify the prompt to refine the extraction logic.

2. Can I add additional columns later?

• Yes, users can update the column list by editing the AI_EXTRACT_TABLE function.

3. Does this feature support all document types?

• Custom table extraction is available for table line item columns only. Click here to learn more about setting key value field prompts

If you need further assistance, contact our support team at [email protected].