The scores provided by the pre-trained Machine Learning Matcher often depend on subjective assessment. Nevertheless, Stibo Systems aims to continuously improve the quality and accuracy of these scores. A dedicated team is ready to engage in a collaborative process to help improve the customer's understanding of the produced scores and to improve the accuracy of the scores through the release of new versions of our pre-trained models.
For detailed information about the Machine Learning Matcher, refer to the Matcher: Machine Learning Matcher topic in the Matching, Linking, and Merging documentation.
If the matching scores you receive result in questions for you or your team, in the Stibo Systems Service Portal, create a ticket with the following details:
-
Summary: Preface your summary content with 'ML Matcher' so it is clear to the support team that the issue relates to the pre-trained Machine Learning Matcher. The format will look like this: 'ML Matcher - <summary of the issue>'.
-
Description: Specify the Model version in use. Provide examples of pairs with current scores and expected scores, and present clear arguments regarding the discrepancies. The examples must include the output of the data elements (normalizers) that are used as input for the Machine Learning Matcher.
-
Issue Type: 'Matching, Linking & Merging'.