LumiParser

Getting Started

How to Increase Extraction Accuracy

LumiParser is highly accurate out of the box, but these steps will help you get the best results for your specific document types.

Use clear, descriptive field names

The AI uses your field names as hints. "vendor_name" works better than "field1". If a field is ambiguous, add a custom AI instruction.

Add AI field instructions

For fields that are frequently wrong, open the field settings and write a short instruction in plain English. For example: "Always return the net amount excluding VAT."

Choose the right field type

Set dates to Date type, monetary values to Amount, and use Classification for fields with a fixed set of possible values. Correct types improve extraction consistency.

Upload high-quality documents

Scanned documents with poor resolution reduce accuracy. Where possible, use digital PDFs rather than scanned images. For scans, ensure 300 DPI or higher.

Use Advanced Reasoning for complex documents

Enable Advanced Reasoning in parser settings for documents with unusual layouts or complex tables. It takes longer but is more accurate on difficult files.

Review and correct early documents

The first few documents you process are the best time to spot patterns in extraction errors. Correcting them in the Inbox helps you identify which field instructions to add.

Was this helpful?