Extracting documents with Sypht

What is document extraction?

Good question! :smiling_face_with_3_hearts: Extraction means a document is being read by an AI product and key-value pairs in the document are identified and tagged in the document. For example, extracting an invoice with the Invoices AI product, things like Amount Due, Amount Paid, Currency code, and so on will be identified and tagged in a document. That’s extraction in a nutshell.

To learn more about the available AI Products, see https://app.sypht.com/marketplace/catalog

How to extract a document:

  1. In the https://app.sypht.com/documents page, select a document and click the Extract button.   Extract__2X.png

  2. In the extraction dialog box, select the AI product(s) you want to extract the document with and then click the Queue button to perform the extraction.

The document is added to the extraction queue and processed. Queue times vary depending on the number of documents being processed.

How to extract tables

Your most important business documents have tables. And it sounds kind of crazy, as tables are already structured, but extracting and structuring data from tables is extremely challenging for AIs. Not a lot of companies can handle them properly. Except for us of course :stuck_out_tongue_winking_eye:!

Sypht has a powerful table tool that allows you to draw and select parts of the table and shows you a preview of the data.

To edit a table

Getting access to table requires you to do the following:

  • Upload the document

  • Extract the document with the proper AI Product

  • Create a validation task for the extracted document.

  • Open the validation task

  • Scroll down to the table, click edit on the Line-item and then click on the table to highlight and edit it.

Sypht-App__2_.gif

 

Adding rows and columns to a table

If the automatic table extraction didn’t match up to exactly what you need, you can modify the table in the following ways:

  • Move existing lines – Hover over a line and then drag it into the proper position

  • Add and remove rows – Hover over the left side of the table and click the -/+ icon to add or remove rows

  • Add and remove columns – Hover over the top side of the table and click the -/+ icon to add or remove columns

  • Clear row – Click the row checkbox and it will deselect the row

table_add_remove_column_row__1_.gif

 

Drawing a table

You can create tables from scratch and then extract the data.

  • Drawing a table – Click the Select table button and outline the table on the page.

  • Add rows and columns – Click on the left side of the drawn table to add and remove rows and

  • Set column categories – Categories define what type of data the column has. Set the column category in the data preview below the table.

drawing_table__1_.gif

 

Moving / deleting a table / setting headers

Tables can also be moved, deleted, and the headers redefined for them. Redefining the header in a table is handy because sometimes a table has a column header instead of a row header.

  • Moving a table – Drag the move icon to rearrange the table.

  • Setting table headers – Click the blue bar at the top of the column so set it as the table header.

  • Deleting tables – Click the trash icon in the field list beside the table to delete the table.

table_move_delete_header__1_.gif

 

Locking tables

You may want to lock your tables from somebody accidentally changing them after you’ve worked so hard to create them. Locking is simple, click the lock icon in the top right corner.

Editing table data

Sometimes the data in the table won’t be quite right, so you can change that with a couple clicks.

To edit a table: Select the cell, click Edit manually, make the change, and then click Confirm.

editing_table_data__1_.gif

 

API extraction

You can extract documents using Sypht’s API, read more about it here: https://sypht.gitbook.io/sypht/upload-and-extract/prediction

Was this article helpful?
0 out of 0 found this helpful