Digital Financial Reporting

The AICPA and others are working on what they call a dynamic audit solution.  For more information see here.  I will find out more.  Here is more information:

• The Future of Audit: Myth or Reality? 
• Caseware
•  Baker Tilly

If you read between the lines of Kalev Leetaru's article Deep Learning Must Move Beyond Cheap Parlor Tricks there is a lot that you can learn.

First, recall that I have pointed out in the past that there are two major techniques for implementing to artificial intelligence:

• Expert systems (logic and rules-based approach): Representing processes or systems using logical rules.
• Machine learning (pattern-based approach): Algorithms find patterns in data and infer rules on their own.

A third approach would be a hybrid of an expert system that also uses machine learning.

Start by reading the concluding paragraph of Leetaru's article:

In the end, for deep learning to move beyond cheap parlor tricks towards solutions that can truly advance society, we must move beyond today’s correlative approaches and simplistic one trick ponies towards algorithms that can actually reason semantically about the world.

That last part "reason semantically about the world" is the key.  That reasoning is powered by things like the functional components defined within ontologies.

Lot's of people are being duped today by software vendors selling smoke and mirrors (i.e. parlor tricks).  I have heard of two projects related to XBRL-based reports and machine learning.  Both were the brainchild of software engineers, not business professionals.  Both were explained to me.  I predicted that both would fail.  I don't know if either succeeded but I doubt it. How do I know this?  Neither involved the hard work of creating the metadata necessary for making the projects a success.

High-quality curated metadata is the answer.  That high-quality curated metadata will supercharge artificial intelligence, including the machine-learning or deep-learning capabilities.  Metadata like this. (This is my best prototype.) Don't misinterpret what you are looking at.  All that human-readable information is also machine-readable.

There are no short cuts folks.  The path that seems cheap and easy might seem seductive, but trust me when I say that the hard, time consuming, expensive path that provides a well engineered, robust system is the only way to get artificial intelligence to work the way we need it to work for things like important functions such as financial reporting.

Don't understand what I am talking about?  I would recommend that you read Computer Empathy, it might help you avoid the snake oil salesmen pushing solutions that will not work.

I have updated the results of my testing of consistency against a conformance suite that I have created for XBRL-based digital financial reporting.

There are now at least three software vendors that provide consistent support for the 11 concept arrangement patternsthat make up the core of XBRL-based digital financial reporting. 

Here is a comparison of the 11 concept arrangment patterns for those three software vendor applications.

Believe it or not, every XBRL-based report submitted to the SEC using US GAAP or IFRS can be broken down into these 11 concept arrangement patterns.  A test set of 6,023 reports which contain 8,532,275 facts can all be arranged into information model that contain these 11 concept arrangement patterns. These and other patterns are the basis for my Financial Report Semantics and Dynamics Theory and the Logical Theory Describing a Business Report.

An ontology is essentially taking the information from those two documents above and representing them in machine-readable form.

I don't know if I should say that something like the US GAAP Financial Report Ontology is the basis for the concept arrangement patterns of if it is the case that the concept arrangement patterns enable the creation of such ontologies.

These same ideas apply not just to US GAAP or IFRS; rather the ideas apply to any financial reporting scheme.  Further, the ideas can be applied to non-financial reporting schemes.

All of this is currently informally standardized using my open source approach. OMG's SBRM is about making my informal standard that I am using for financial reporting more formal plus it is about expanding the ideas created specifically for financial reporting and applying them to the more general use case of general business reporting.

Jigs