Key-Value Field Extraction Using LLM Prompts

Overview

With LLM-Powered Key-Value Field Extraction, users can enhance the accuracy of data extraction by providing prompts directly in the Description field. This allows the AI model to better understand the expected output format and improve field recognition.

Feature Highlights

• AI-Guided Extraction: Users can specify extraction logic for each field using natural language prompts.
• Customizable Data Fields: Define expected values for each key-value pair to improve accuracy.
• Seamless Integration: Works with existing document types for enhanced extraction capabilities.

How to Use Key-Value Field LLM Extraction

Step 1: Navigate to Field Settings

1. Open the Edit Fields section for your document type.
2. Select the field you want to configure (e.g., Invoice Number).

Step 2: Define the Extraction Prompt

1. Under the Basic tab, locate the Description field.
2. Enter a natural language prompt describing how the AI should extract the data.

Example Prompt:

For extracting an Invoice Number, you can use:

Extract the invoice number from the document. The value should be alphanumeric and typically found near the header or invoice title.

For extracting a Due Date, you can use:

Find the due date of the invoice in the format DD/MM/YYYY. The date is usually labeled as "Due Date" or "Payment Due" in the document.

Step 3: Save & Apply

• Click Save to apply the prompt to the field.
• The AI model will now use this prompt to extract values more accurately when processing documents.

FAQs

1. What if the extracted value is incorrect?

• Try refining the prompt with more details or examples.

2. Can I use this for all document fields?

• Yes, any field can have a custom prompt to improve extraction accuracy.

3. Does this work for numeric or date fields?

• Yes, you can specify formats in your prompt (e.g., "Extract the date in DD/MM/YYYY format").

For further assistance, contact our support team at [email protected].