form recogniser

Question

form recogniser

Sivagnanam, Sinduja (Cognizant) 6

When testing custom model with multi page document, the result does not extract all rows from the table. Why certain values in table rows are not extracted? But those are still recognized by OCR. How to improve the extracted outcome?

Rohit Mungi 49,131 Reputation points Microsoft Employee Moderator

2022-06-23T11:37:13.777+00:00

@Sivagnanam, Sinduja (Cognizant) Does the table span multiple pages in a document? The custom model does not support tables that span multiple pages.
If the table does not span multiple pages but if it still fails, could you try to check if layout API works with the same document?
Sivagnanam, Sinduja (Cognizant) 6 Reputation points

2022-06-23T12:10:35.213+00:00

multi page tables are supported with 2022-06-30 api version and the model is extracting the table data across the pages, but not all values. For, eg. certain values in row 3 , row 4, row 10 etc are not getting extracted. Also for some column it seems to combine values from previous or next row. how to ensure all values for all rows are extracted and only the data from relevant rows are being extracted. Layout could not be considered as we need labelled and selected values.