Document Parser not recognizing table boundaries

[emily_ross]

I’m parsing insurance claim forms and the Document Parser is having trouble with tables that don’t have visible borders. The data is clearly in a tabular layout but the parser treats each cell as a separate text block instead of recognizing the table structure.

Has anyone dealt with this? Is there a setting to adjust table detection sensitivity?

Example: the form has columns for Date of Service | Provider | Charge | Status but they’re just aligned with spacing, no grid lines.

[dev_kumar]

I ran into something similar. What worked for me was pre-processing the PDF to add faint grid lines using a script before sending to the parser. Hacky but effective.

Alternatively, if you’re using the API you can pass table_detection: "aggressive" in the parse options and it does a better job with borderless tables.

[emily_ross]

The aggressive table detection option did the trick! Didn’t know about that parameter. Thanks Dev!