Train multi-row tables

This quick feature allows the model to accurately extract information from tables spanning across multiple rows.

Procedure

  1. Train the first row of the table.
    Visual to train the first row
  2. Click Add more rows icon on the first row.
    Add a new row
  3. Draw a bounding box on subsequent rows or columns within the table. By doing so, the subsequent nested entities will be automatically detected and identified under each parent.
    Draw a bounding box
  4. Review the identified text values to verify that they accurately represent the content of the document.
    For detailed information, see the Label a document section in Create a Custom Document Extractor.

Next steps