xSuite Helix Capture extracts document content using various methods (traditional and AI-based) and also incorporates master data. The system evaluates these data sources comparatively to ensure that the most reliable result is used for the subsequent process.

With this release, we have enhanced the extraction process to include this mathematical comparability—that is, a normalized calculation of confidence scores. Normalization ensures that all extraction results are evaluated on the same basis. This allows the system to reliably compare results internally and provides the user with a more reliable basis for validation. Additional data sources, such as master data, can be incorporated into the quality assessment of the data.

The normalized confidence scores fall within a fixed range between 0 and 1. A score of 0 indicates invalid or implausible field content, while a score of 1 indicates externally verified and therefore highly reliable field content.

As part of the data validation process, the system can, for example, increase the confidence score of an extracted field value if that field value also appears in master data tables.

If a corresponding entry is missing from the master data, the system can reduce the score to 0. In addition to the master data, xSuite Helix also takes into account configurable rules for calculating and internally comparing confidence scores. For example, you can specify that a document date must not be older than one year. You can flexibly adapt these rules to project-specific requirements.

Based on this, during validation, the system suggests the field content with the highest confidence score for each field. In addition, xSuite Helix offers alternative field values provided that their calculated confidence scores exceed a defined minimum value.

As part of a data validation project, the system derives the information required for further processing in the connected ERP system from the read and validated field contents. This includes, for example, the company code, the vendor ID, or the tax code.

Until now, xSuite Helix has distinguished between different types of tax identification numbers during vendor recognition. In Germany, for example, this primarily involved differentiating between the VAT identification number and the tax ID number. This distinction was made using the " Tax Type" field in the sys_BusinessPartner master data table.

However, since ERP systems use different terms to refer to tax types, xSuite Helix no longer takes this information into account.

During vendor recognition, xSuite Helix compares the tax ID numbers on the document with all tax ID numbers stored in the master data. The original classification no longer plays a role in this process.

The advantage is that vendor recognition continues to work even if you mix up or confuse different types of tax ID numbers. You can also use other official registration numbers (e.g., the "Company Registration Number" in the UK) for vendor recognition.

To completely exclude a selection criterion (such as the purchase order number) from vendor recognition, you must now deactivate the selection criterion in both the extraction project settings and the validation project settings.

The settings of an extraction project only have an influence on the processes of classic document extraction. The configuration does not affect document segmentation, processing by xSuite AI, or downstream data validation. These components use their own configuration logic. The project configuration for classic extraction has no effect on the components mentioned.

Therefore, disable selection criteria—such as bank details, tax ID numbers, and email or web addresses—directly within the respective validation project. At this level, you control which criteria should be taken into account for vendor recognition.