PARSER GUIDES
Troubleshoot Extraction Accuracy
Common Problems
Fix missing, incorrect, or misplaced extracted values
Start by checking whether the problem affects one document or repeats across several documents:
- One document: Check the source file and correct the value manually.
- Several documents: Improve the parser field, table, or document-type setup, then test again.
A value is wrong in one document
1. Open the document.
2. Compare the extracted value with the original page.
3. Click the incorrect value, enter the correction, and save it.
If the original text is blurry, cut off, or covered, upload a clearer version when possible.
See Check and Correct Extracted Data for the editing steps.
The same field is often wrong
1. Open Settings -> Fields & Tables.
2. Give the field a specific name, such as Total Including Tax instead of Total.
3. Add an AI Description explaining exactly which value to capture.
4. Save the fields.
5. Re-process two or three test documents and compare the results.
Describe how to distinguish the value from similar information. For example: "Use the purchase order number supplied by the customer, not the invoice number."
A field is empty
Check these questions in order:
1. Is the value visible in the original document?
2. Does the field name clearly describe it?
3. Would an AI Description remove ambiguity?
4. Is the selected document type suitable for this file?
Enter the value manually when you need to complete the current document. Change the parser setup when the problem repeats.
A value appears in the wrong field
Use distinct names and descriptions for the two fields.
For example:
- Invoice Date - The date the supplier issued the invoice
- Due Date - The date payment is due
After saving the changes, re-process a small test set before using the parser for more documents.
Table rows or columns are wrong
Check that:
- Repeating information is configured as a Table
- Every column has a specific name and the correct field type
- Similar columns, such as Unit Price and Line Total, have clear descriptions
- The source document shows readable row and column boundaries
For long or irregular tables, follow Table Extraction Tips.
Most fields are wrong
This usually points to a broader setup problem:
- The parser uses the wrong Document Processing Type
- Different document types are being sent to one parser
- The source file is difficult to read
- The parser fields do not match the information in the documents
Compare the setup with Choosing a Document Type, then test one representative file again.
Before re-processing documents
Re-processing uses the parser's current fields and settings and may use credits. Test the changes on a few documents before re-processing a large batch.
See Re-processing Documents for the available actions and credit considerations.
If the problem continues
Record the parser name, document name, affected field, expected value, and extracted value. Share those details with your team administrator or PARSER support.
Next steps
- How to Increase Extraction Accuracy - Prevent common accuracy problems
- Add and Edit Fields - Improve field names and descriptions
- Confidence Scores - Identify values that may need review