JSON clean up from multiple PDF pages.

KNIME Analytics Platform
,
1

Using azure form recognizer to read PDFs for some data capture.SO after a few trials, I got a clean CSV extract from the JSON file. so far so easy. The problem is with PDFs you can get 2 or 3 tables on a single page, and as they have row and index numbers… if you try to pivot the results from a multi-page PDF you get a mess.

JSON_workflow
JSON_workflow1008×630 10.1 KB

So my new workflow splits every table into a clean “row”. So my JSON to table output (the last step) is ok but ideally, I need to loop through this output and clean it a bit. Because I need to PIVOT on all 8 rows. For each row, i group on “row index” and PIVOT on columnIndex and aggregation is “text”.
SO basically the output should be 14 separate CSV tables…
I have uploaded my workflow and the source file. (JSON)Layout-Result-Pages from 2018 AR RIO (2).json (1.5 MB)
node 56 and node 59
node 56 and node 591008×630 15.9 KB
JSON to Table
JSON to Table1109×702 26.5 KB
UNGROUP_2
UNGROUP_21008×630 18 KB
JSON_PATH_2
JSON_PATH_21008×682 23.2 KB
UNGROUP_1
UNGROUP_11008×630 17.6 KB
JSON_PATH_1
JSON_PATH_11008×667 22.6 KB
JSON READER
JSON READER1008×630 10.1 KB
JSON to Table
JSON to Table1008×630 23.1 KB
2021-02-01T00:00:00Z

2

Hey @bruce_231,

I will have a look and get back to you shortly!

Best,
Julian

1 Like
3

HI Julian
thanks for picking this up! I actually figured out a second “part” of the workflow that solves the issues. My new problem is how to connect both workflows… Create the JSON files (done), transform the JSON files to CSV (done)… but no connected, I raised another post :slight_smile:
And I’m looking at the suggestion (a call workflow node)…

2 Likes
4

This topic was automatically closed 182 days after the last reply. New replies are no longer allowed.