3749597 - CURRENCY_NOT_RECOGNIZED Error Due to Inferred Currency in Document Extraction

Image/data in this KBA is from SAP internal systems, sample data, or demo systems. Any resemblance to real data is purely coincidental.

• Payment advice documents do not contain currency information, but the extraction result returns a currency.
• Example (the following examples use dummy data for illustration purposes only) :
  • Case 123 → Extracted currency: ZAR
  • Case 321 → Extracted currency: EUR
• Extraction response shows error:
  "lastActionStatuses": [
    {
      "status": "FAILED",
      "messages": [
        {
          "code": "CURRENCY_NOT_RECOGNIZED",
          "message": "Currency code 'ZAR' was not properly recognized (coordinates are 0.0)",
          "type": "ERROR",
          "doxUIRelevant": true
        }
      ]
    }
  ]
• The extracted currency (e.g. ZAR – South African Rand) is not enabled in the customer system. 

• SAP Service Cloud Version 2 1.0
• SAP Sales Cloud Version 2 1.0
• Document Information Extraction (Business Document Processing – Payment Advice)

1. Upload a payment advice document that contains no printed currency code.
2. Run extraction and observe that a currency value is returned in the result despite no currency in the source.

• The Document Information Extraction service is based on machine learning (ML).
• The service may:
  • Infer values based on document structure, learned patterns, or contextual signals
  • Return values even when they are not explicitly present in the document
• When a value is inferred:
  • It may not have valid positional data
  • This is reflected as:
    • Coordinates = (0.0, 0.0)
• If such a value (e.g. currency) is:
  • Not configured in the system, or
  • Not expected by business logic
  • → The system raises:
    • CURRENCY_NOT_RECOGNIZED

• System Behavior
  • This behavior is working as designed.
  • The extraction service may return inferred values even if the document does not explicitly contain them.
• Workaround
  • Exclude fields with:
    • Coordinates = (0.0, 0.0)
  • Optionally:
    • Manually override values in UI if required
• Recommended Approach:
  1. Validate Extracted Fields Using Metadata
     • Each extracted field includes metadata such as:
       • Coordinates
       • Confidence score
     • Fields with:
       • Coordinates = (0.0, 0.0)
       • indicate:
         • The value was not detected from a physical position in the document
         • Likely inferred by the ML model
  2. Access Extraction Results from Integration Payload
     • According to SAP Help documentation for Document Extraction:
       • The extraction results are provided as part of the processing output
       • In integration scenarios (e.g. iFlow):
         • The payload includes an attachment containing extraction details
         • The exact structure and naming of attachments depend on implementation.
       • This attachment:
         • Contains extracted values along with metadata (e.g. coordinates, confidence)
     • Clarification:
       The exact attachment naming convention (e.g. *extraction_results.json) and structure can vary depending on the integration implementation and configuration.
       However, SAP documentation confirms that detailed extraction results including metadata are available for downstream processing. 
  3. Implement Filtering Logic in Integration Layer
     • In your middleware (e.g. SAP Integration Suite iFlow):
       • Parse the extraction results payload
     • Apply filtering rules such as:
       • Ignore fields where:
         • Coordinates = (0.0, 0.0)
       • Optionally:
         • Apply confidence threshold filtering

           This ensures:

           • Only physically detected (reliable) fields are processed
           • Incorrect inferred values (e.g. unwanted currencies) are excluded

• Configure Business Document Extraction Schemas
• Business Document Integration

payment advice, currency extraction, currencyCode, coordinates 0.0, dox ai, document ai, esm business document service, CURRENCY_NOT_RECOGNIZED, low confidence, post-processing, confidence threshold, zeroed coordinates, inferred currency, payment advice excel, business document extraction, ML, CA-ML-BDP-COE, DOX AI: PAYMENT_ADV  , KBA , CEC-CRM-BUSD , ESM Business Document Service , CA-ML-BDP-COE , ML models of business document processing services , Problem