when using a Custom Extractor-Document AI, some extracted fields appear with inverted word order.
Examples:
“MARIA ISABEL” —> “ISABEL MARIA“
Expected behavior
The ocr output should preserve the original visual word order as it appears in the document
Hi @Isabel_Sanchez_Benit
Welcome to the AI Forum!!!
Thanks for reaching out!
This problem usually occurs when the labeling sequence during model training does not align with the visual reading order of the document.
Please check your training data in the Document AI Workbench, make sure the tokens are selected in the correct left-to-right order, and retrain the model.