In my latest YouTube recording, released at the end of last week, I addressed three primary topics:
- Templated versus Template-less Intelligent Document Processing (IDP).
- Operator-Based Training.
- Line Item Matching.
IDP tools should have evolved beyond the use of templates to recognise correct information on a page, and should now be moving towards AI, Machine Learning, and Entity Recognition. A templated-only approach implies the following:
- Every company that currently sends you documents would require a custom template, leading to substantial setup costs.
- Each new company that sends you a document will necessitate its own custom template, leading to ongoing maintenance costs.
- If a company alters their layout, they will require a new custom template.
- Templates can typically only be created and modified by an administrator or developer.
An efficient template-less solution should also enable the user to improve recognition results on the fly using point-and-click techniques, often referred to as Operator-Based Training. This training should be applicable to both static and table-based fields.
IDP solutions should also facilitate matching data from the document back to the data stored within the ERP system. This could include:
- Identifying customers, suppliers, delivery addresses based on names, addresses, VAT registration numbers, bank account details already held within an existing dataset.
- Matching document references such as order numbers, delivery numbers, account codes found within an existing dataset.
- Matching line items from the document to existing datasets such as order lines and delivery lines.
The video demonstrates these techniques in use with a live example in Abbyy FlexiCapture. Here is the link: https://youtu.be/k0XhBWzWqow.