How Do You Correct Incorrectly Extracted Data in Whitevision?
Quick Summary
Whitevision B.V. Declaraties is a Dutch document-processing platform that uses OCR and artificial intelligence to read invoices, receipts, and other business documents automatically, but even a high-performing recognition engine occasionally misreads a field. Consequently, knowing how to correct incorrectly extracted data quickly and properly matters just as much as understanding how extraction works in the first place, since a well-made correction not only fixes the immediate document but also teaches the system to avoid the same mistake in the future. This article reveals how to correct incorrectly extracted data in Whitevision, covering why errors occur, how to fix different types of misread fields, how the built-in feedback function trains the system, and how to prevent recurring issues from a specific supplier or document type.
Why Does Extracted Data Sometimes Need Correction?
What Typically Causes a Recognition Error?
While Whitevision’s OCR and AI engine generally achieves a high recognition rate from the outset, certain conditions make errors more likely. These commonly include:
| Cause | Example |
|---|---|
| Poor image quality | A blurry photo or a faded thermal receipt |
| Unusual document layout | A new supplier with a non-standard invoice format |
| Handwritten elements | A handwritten note or amendment on a receipt |
| Ambiguous data | Multiple similar amounts appearing close together on a document |
Because these situations are relatively predictable, understanding them in advance can help staff anticipate when a document is more likely to require a correction.
Why Is It Worth Taking Corrections Seriously?
Since Whitevision’s recognition engine improves through a feedback function that learns from user corrections, every fix applied during processing does more than resolve a single document; it also refines how future documents of a similar type are read. Therefore, treating corrections as a quick, careless task rather than a genuine input to the system’s learning process can slow down the overall improvement in recognition accuracy over time.
How Do You Identify That Extracted Data Needs Correction?
What Are the Warning Signs of a Misread Field?
Before correcting anything, it helps to recognize the signs that a field has likely been misread. Common indicators include:
- A total amount that does not visually match the figure printed on the document
- A supplier name that appears garbled, incomplete, or entirely incorrect
- A date formatted in a way that clearly does not correspond to the document
- VAT figures that do not add up correctly against the net and gross amounts
Since these issues are usually visible at a glance when comparing the extracted data to the document image shown alongside it, a brief visual check before approval is generally enough to catch the majority of errors.
How Does the Review Screen Help You Spot These Issues?
Because Whitevision displays the recognized header and line data next to the original document image on the same screen, spotting a discrepancy typically does not require any special tools or a separate verification step. Consequently, most corrections are identified naturally during the standard review process, rather than requiring a dedicated audit afterward.
How Do You Correct a Misread Header Field?
What Are the Steps to Fix an Incorrect Header Value?
When a header field, such as the invoice number, supplier name, or total amount, has been misread, the general correction process involves:
- Locating the incorrect field on the review screen.
- Comparing it against the corresponding value shown in the original document image.
- Editing the field directly to reflect the accurate value.
- Confirming that any related fields, such as VAT calculations tied to a corrected amount, update consistently.
- Saving the correction before proceeding to approval or booking.
Afterward, it’s worth briefly checking that the corrected value now matches the document exactly. This matters because a rushed correction can sometimes introduce a new error rather than resolving the original one.
How Do You Correct Supplier or Bank Account Details?
Incorrect supplier or bank account details carry a higher financial risk than most other fields. Because of this, correcting these values deserves particular care. If a supplier’s name or account number appears incorrect, it’s advisable to cross-reference it against existing supplier master data in your financial system before saving the correction. This helps confirm the fix is accurate, rather than simply different from the original misread value.
How Do You Correct Line-Item and VAT Data?
How Do You Fix an Incorrect Line Item?
For documents with multiple line items, correcting a single misread line typically involves a few steps. First, select the specific line. Then adjust the relevant value, such as quantity, unit price, or VAT rate. Finally, confirm that the line total recalculates correctly. Line-item errors are more common in invoices with dense or unusually formatted tables. Because of this, it’s generally worth reviewing line data more closely for these types of documents than for simpler ones.
How Do You Handle Recurring VAT Recalculation Issues?
If VAT is consistently miscalculated for a particular type of document, this often points to a deeper configuration issue rather than a one-off recognition error. In this case, it is advisable to:
- Confirm the VAT rate configured for the relevant product or service category is accurate
- Check whether the supplier’s invoices consistently follow a layout that the system struggles to interpret
- Escalate the pattern to an administrator so that a dedicated rule can be configured for that document type
Consequently, addressing a recurring VAT issue at the configuration level tends to be more effective than correcting the same type of error manually on every invoice.
How Does Whitevision Learn from Corrections?
What Is the Feedback Function and How Does It Work?
Whitevision includes a feedback function that allows the system to learn automatically from corrections made during processing. Once a correction is made and, where applicable, the feedback option is activated, the system remembers the adjustment and applies it automatically the next time a similar document is processed. As a result, a correction that initially takes a few seconds to make manually often needs to be made only once for a given supplier or document type.
How Does This Benefit the Wider User Community?
Because Whitevision’s recognition engine benefits from collective learning across its user base, corrections contribute not only to an individual organization’s future accuracy but also to the broader knowledge the system draws upon. Therefore, even a new document type introduced to your organization may already benefit from recognition improvements learned from other users processing similar documents elsewhere.
How Do You Prevent Recurring Extraction Errors?
What Should You Do If the Same Error Keeps Appearing?
If a specific field continues to be misread despite repeated corrections, this typically points to one of a few causes. The initial correction may not have been saved properly. The document format may vary too much for a single learned pattern to apply. Or a deeper configuration adjustment may be needed. In these cases, it’s generally best to escalate the issue to an administrator rather than continuing to apply the same manual fix repeatedly.
How Should You Escalate a Persistent Recognition Issue?
To escalate a recurring issue effectively, it helps to:
| Step | Purpose |
|---|---|
| Document the specific error | Provides a clear example for troubleshooting |
| Note the affected supplier or document type | Helps narrow down the root cause |
| Confirm whether the issue is isolated or widespread | Determines if it affects one user or the whole team |
| Contact Whitevision support if configuration changes are needed | Ensures a lasting fix rather than a repeated manual correction |
Since Whitevision support can access tools such as remote assistance to investigate persistent issues directly, escalating a well-documented, recurring problem is generally more effective than attempting to resolve it through repeated manual corrections alone.
Conclusions
Correcting incorrectly extracted data in Whitevision B.V. Declaraties is a quick, straightforward process made easier by the platform’s side-by-side review screen, but it carries importance well beyond the individual document being fixed. Because the built-in feedback function allows the system to learn from every correction, addressing errors properly, rather than working around them, steadily improves recognition accuracy for future documents from the same supplier or of the same type. By recognizing the common causes of misreads, correcting fields carefully, and escalating persistent issues when they arise, organizations using Whitevision B.V. Declaraties can ensure that occasional recognition errors become progressively rarer rather than a recurring source of manual work.
Frequently Asked Questions
Generally, corrections should focus on values that are factually incorrect, such as a wrong amount or date, rather than minor formatting differences that do not affect the underlying data. However, if a formatting inconsistency causes issues further down the process, such as in ERP field mapping, it is worth flagging even if the value itself appears correct.
If a correction does not seem to be sticking for a specific supplier, this may indicate that the feedback function was not properly activated when the correction was made, or that the supplier’s invoice layout varies enough between documents that the system has not yet learned a consistent pattern. Reviewing this with an administrator can help determine which explanation applies.
In most cases, an incorrect correction can be fixed by simply editing the field again with the accurate value, since this follows the same correction process described above. However, if an inaccurate correction has already influenced how the system recognizes similar future documents, it is advisable to inform an administrator so they can confirm the learned pattern is adjusted accordingly.
