Using Metadata for Document based Data Capture

Let’s consider a video film. It’s a compilation of many static picture frames, each one capturing snapshot of the action at a time instance. A collection of such frames when viewed at a certain speed delivers a sensible outcome to the viewer. Combined with the associated audio, it becomes complete and further enhances the experience. The same cannot be experienced by viewing each frame independently.

Picture source: IEEE Paper - Shlizerman_Audio_to_Body_CVPR_2018_paper.pdf

This sense is made possible by the relative sequencing of each frame with the next logical succeeding one. The common factor contributing to this sense is the metadata of each frame. This metadata carries the attributes of each frame and enables creating a fabric of sequential relativity of the data itself in conjunction with the preceding and succeeding frames.

In a process involving data capture from documents, the focus is primarily to capture data from each document independently. Arguably, the use of metadata of the captured data fields is limited. It should be noted that depending on the industry and the documents, some fields occur only once while several others occur in multiple places either within the same document or across different other document(s). This redundancy, if tapped appropriately, has potential to improve the accuracy of data capture and thus improve the Straight Through Processing (STP).

The nature of such data fields does not vary much regardless of the document they appear in. The opportunity to validate them depends on how the metadata of these documents is utilized. An enterprise building smart automation solutions for data capture can further enhance its capabilities by leveraging the metadata fabric. 

For example: the US residential mortgage industry sources, reviews and processes 100+ document types across the life-cycle.

Although the industry bodies have provisioned certain standard templates, across 3000+ counties there are many localized variations in use. And, basis the parties involved in the transaction, the documents commonly required relate to the below. There may be other additional documents based on the nuances.

1.      Parties involved

a.      Borrower(s) – individual (partners, married, divorced, etc.) / institution

b.      Lender – individual / institution

c.      Trustor / Escrow agency

2.      Property particulars – residential/commercial, appraisal / valuation report, taxes paid/due, hazard insurance, litigations, etc.

3.      Credit rating report

4.      Assets & debts (income tax returns, pay stubs, alimony, etc.)

5.      ID – citizen / alien 

Pictures source: Google for Deed of Trust, Employment Application, 1004 and Experian for sample credit report

A simple illustration of commonly used fields and the options to cross-validate them are as given below: