Table Recognition Training - Cell Text Tokens in annotations

[johntoma]

Do these cell text tokens (highlighted in the image below) in the training data annotation have any affect on table recognition training?

Is it necessary to include these text values, or can I just leave it blank?

I am confused because the OCR for table recognition is done by a separate model, so I'm not sure if these tokens are important or not. I assume they are not actually necessary for training at all and have no affect on training results and are only there for human readability?

Many of the tables in my training data either don't include these tokens (the "tokens" array is simply left empty in the annotation) or have the incorrect tokens for that cell.

Do I need to review my dataset again and ensure these cell text tokens have the correct values, or am I correct in thinking that these token values are not important?

Thanks for your time.

[github-actions]

This issue is stale because it has been open for 90 days with no activity.