tables and rows confidence in test

Hi Tamir Bar-Tov,
Greetings & Welcome to Microsoft Q&A forum! Thanks for posting your query!

In Azure Document Intelligence, the confidence for tables and rows in the test set is calculated based on how well the model's predictions align with the labeled ground truth data provided during training. The quality and consistency of the training data play a critical role in determining the confidence scores, as the model learns patterns and structures from it. During testing, the model evaluates table boundaries, row structures, and cell alignments by comparing them with the labeled data. It checks for structural accuracy, text placement, and handling of complex layouts like merged cells or closely spaced rows. Confidence is influenced by the completeness and accuracy of the training labels, as well as the model's ability to generalize patterns learned from the training set to the test data.

I hope this information helps. Thank you!

Hi Tamir Bar-Tov,
Following up to check if the above suggestion was helpful. If you have any further queries, please feel free to let us know.
Thank you!